Skip to main content

Full text of "Grammatical Literature"

See other formats

f-!^dAA!:4iv!‘:i*‘*\r!/'‘\t\Si':^S‘t !'‘‘-'<>rt/’'''^\ ■''Ae •.■'{'•••".■f'., i' •'"' , . .'‘;i ' ' '-■'■'v.. ‘j/i ' , r'„\' 









Ease. 2 








Contents of Vol, V 

Vol. V : Scientific and Technical Literature^ Part II 

Fasc. 1 : J. D . M. Derrett 

Fasc» 2: H. Scharfe 

Fasc. 3; E. Gerow 

M. Hahn 
C. Vogel 

Dharmasastra and Juridical Literatuxe 
Grammatical Literature 
Indian Poetics 


CIP-Kurztitelaufuahme der Deutschen Bibliothek 
A history of Indian literature / ed. by Jan Gonda 
KE: Gonda , Jan [Hrag.] 

Vol. 5. ScientiSc and technical literature : P. 2. Fasc. 2.-. Scharfe, Hartmut : Grammatical literature 

Scharfe , Hartmut 
Grammatical literature. 

technical literature; Paso. 2) 

vortahalten. Photographlsche und photomechanische 
burg. Printed in Germany. Sigei : H I L. Verlages. Gesamtherstellung : Friedrich Pustet, Pegens- 


Chaptee I The Origins 77 

Chaptee II Panini 88 

Chaptee III Yaska 117 

Chaptee IV Shadows of Some Early Theorists 124 

Chaptee V The Pratisakhya-s 127 

Chaptee VI Katyayana 135 

Chaptee VII Fragments Preserved in the Mahabhasya 149 

Chaptee VIII Patahjali 152 

Chaptee IX The Buddhist Sanskrit Grammarians 162 

Chaptee X The Jain Sanskrit Grammarians 168 

Chaptee XI The Later Panini School 170 

Chaptee Xn The Siksa-s 176 

Chaptee XTTT Grammars of the Dra vidian Languages . 178 

Chaptee XIV Other Systems of Sanskrit Grammar 187 

Chaptee XV Grammars of the IVIiddle Indo-Aryan Dialects . . . . 191 

Chaptee XVI The ParasLprakasa 196 

Chaptee XVII Grammars of the New Indo-Aryan Languages . . . . 198 

Abbreviations 200 

A Selected Bibliography . . . ' 202 

Index 210 

Hartmut Scharfe 


Chapter I 


The power of speech/language (vac) has intrigued Indian thinkers from the 
earliest times we know of.^ Words were not merely the poet’s tools, not only 
the magic keys by which the officiating priest opened the door to prosperity 
and heavenly bliss: often Speech was seen as a causal force behind even the 
gods and the universe. The Vedic verses dealing with Speech are couched in 
expressions as dark and mysterious as the power they praise: “Three parts 
which are hidden, mortals do not activate; the fourth part they speak” gwJid 
trlni ['padd7ii] nihitd nemgayanti tuny am vdco manusyd vadanti (Egveda I 164. 
45cb).2 When Speech was visualized as the cosmic Cow, her steps (pada) were 
first taken as the lines of the verse {catuspadd tristubh Tait. S. Ill 2, 9, 1); 
advanced analysis saw in her steps the single words (Sat. Br. X 2, 6, 13^), and the 
(usually four) lines of a verse were henceforth called the Teet’ (pdda; e.g.. Ait. 
Br. IV 4, cf. Greek Aksara, origmally perhaps ‘the unmoving’ part of 

the fiow (yksar) of speech, came to denote the smallest element of speech rec- 
ognized at that time, i.e. the syllable whose importance for the Vedic priests 
lay in the syllable counting nature of their metres the term became so popular 

1 0. Strauss, ZDMG 81.99-105; L. Eenou, Etudes vediques et panin^ennes I 
(Baris, 1955), 1-27; A. Padoux, Eecherches sur la symbolique et T^nergie de la 
parole dans certains textes tantriques (Paris, 1963), 15-40; W.N. Brown, inPrati- 
danam (Fs. F.B. J. Kuiper, The Hague, 1968), 393-397. 

2 A very original interpretation of this verse is Sat. Br. IV 1, 3, 16: one fourth of 
speech each for men, beasts, birds and small vermin. Note also Chand. Up. I 1,2 
purusasya vdg rasah “the essence of man is speech.” 

3 B. Liebich, Zur Einfuhrung in die indische einheimische Sprachwissenschafb II 
(Heidelberg, 1919), p. 5; slightly different L. Ebnou, JA 233. 134-138. 

4 J.A.B. VAN Buitenen, JAOS 79.176-187. 


H. Scharfe • Grammatical Literature 

that we find it even in a rare colloquial form ahhJchallkrtya ‘syllabizing’ in Rgveda 
VII 103, 3.3 The mystical speculation on the power of speech and the efficacy 
of a properly chanted mantra is well documented in later Vedic texts; its heirs 
are probably the tantric systems attested since the middle ages but almost cer- 
tainly much older than that.® 

The interest in language deepened in the Brahmana period. Important words 
were frequently explained by relating them to a verb: yad . . . aichams, tad 
istindm istitvam “because they wished to . . the offerings are called isti’’ (Ait. 
Br, 1 2, 1). This etymology is ‘false’ because isti ‘offering’ is historically derived 
from i/ya/ ‘worship’ and not from j/^’s ‘desire’ (from which there is a homonym 
isti ‘desire’); but it underscores an aspect of every offering.’ The etymology 
helps to fathom the full and ‘real’ meaning of a noun, but does not necessarily 
represent the ‘grammatical’ conviction of the author.® In time, this kind of 
etymology was to develop into a special auxiliary science (veddnga) nirukta 

The first branch of linguistics to attain independent status was the study of 
phonetics. The attempt to preserve the sacred texts in a strictly oral tradition 
(and to preserve not only the words but also their correct pronxmciation!) led to 
an inquiry into the production of the sounds of speech. A few terms like varna 
‘sound’ or avasdna ‘pause’ appear in the earliest Brahmana-s, and at the time of 
the Aranyaka-s and Upanisad-s the science was probably fully developed (vide 
Ait. Ar. in 5-6, Tait. Ar. VII, Chand. Up. II 22, 3-5). The alphabet (aksara- 
samdmndya) consisted of vowels (svara), stops (sfarsa)^ semivowels (antastha 
for *antas~stJui) and spirants ( u^man). One had recognized the role of voicing® 
and of mouth aperture, the nasalization of vowels and the nature of the stops 
as contact sounds as the name sparsa lit. ‘contact’ shows. Single sounds were 
named with an attached element -kdra ‘maker’ in analogy to older expressions 
like hin-kdra ‘the sound hin^ (which itself is based on still older verbal hinn akrnot 
^made Am’ Bgveda 1 164.28) or vasat-kdra ‘the exclamation vasat''^^: a-kara 
/a/’, im-kdra ‘/n/’ (consonant names employ an extra /a/ to facilitate pronun- 
ciation).^^ It is not surprising that the new science was usually referred to as 
just the study’ (slksd^ later siksd}^)\ its categories were fundamental for aU 
further linguistic studies as was its pure interest in sounds rather than letters. 

s P. ThiemEjKZ 71.109. 

fi A. Pabotts:, Becherches, p. 43 f. 

Cf. also Atharva-veda III 13, 1—4 with four verb etymologies, 

® Th. Aufbecht, The Aitareya Brahmana (Bonn, 1879), p. 432. 

« It be noted however that the technical side of voicing was not under- 

^ Indians never knew about the vocal cords: P. Thxeme, ZDMG 
1U7. boo. 

Of. also the philosophical term ahamkara ‘the exclamation aham “I”’* JAB 
VAiT Buttekek, JAOS 77.17. * 

L.Rekotj, JA233.149f. 

12 H. Ludebs, Die Vyasa-giksha (Kiel, 1895), p. 1, fn. 1. 

The Origins 


It was taught in six chapters: varim ‘sound,’ svara ‘accent,’ mdtrd ‘quantity, 
bala ‘articulation,’ sdman ‘recital’ and suTritdna ‘connection. 

There can be no doubt however that these insights were in due time applied 
to the art of writing. According to the dominant theory the Brahmi script goes 
back to a North Semitic script as it was used in the eighth to sixth century B. 0. 
according to others the source may have been a South Semitic script; and still 
others believe that the script is autochthonous. Whatever the origin, the pho- 
netic principles of the iiJcsd have been applied to this script: a separate letter 
(and only a single letter!) denotes each phoneme; short and long vowels are 
differentiated.^^®, alphabet^® is phonetically ordered which raises it above 
the Semitic (and European) arbitrary arrangements. The same principles charac- 
terize various later scripts found in India and in neighbouring countries where 
Indian culture traveled. 

A distant and unexpected echo of this script (and indirectly of the phoneti- 
cians’ work) is the rise of Chinese phonetical science in the fifth or sixth century 
A. D. with the creation of an almost phonetical script for Chinese. By the so-called 
fan^-chHeh^ principle original word signs are used to denote initial and final 
consonants alone, e.g. Jcu^ plus san^ together denote hari^; the inspirational 
source was the Indian siddha-sciipt used for charms by the Chinese Buddhists-^*^ 
The Korean ^nmun-^cn^t (announced by royal proclamation in the year A. D. 
1446 ; note the alphabetical order: k kh ng t th n p ph m . . .) is a farther (maybe 
indirect) reflexes as also the disputed A Aim-script of Japan.i® In the west, In- 
dian influence on the Arabic grammarian (eighth century A.D.) at Basra 

IsTote that the metrical schools also use mdtra but with different values. For a 
phonetician/grammarian a short vowel has 1 mdtrd, a long vowel 2 mdtrd-a and a 
consonant 1/2 mdtrd. In metrics an open syllable with a short vowel has 1 mdtrd, 
a closed syllable and a syllable with long vowel 2 mdtrd-s. The length of agdt would 
be phonetically 1 +V 2 +2 -fV 2 =4 mdtrd-B, metrically 1 -f 2 = 3 mdtrd-s. 

1^ Tait. At., ch. VII. sarntdna is probably the later sarndhi. The discovery of 
sandhi was utilized early in mystical speculation ; the sacred syllable om is allegedly 
made up of three elements: a -f w +m > orri (Ait. Br. V 32, 2). 

15 H. Jensen, Sign, Symbol, and Script, 3rd ed. (New York, 1969), p. 369. 

15a The step from a half syllabic sign to single letters was taken only in rare ca- 
ses: Mahanisiha ed. W. Sghubeing, APAW 1918, p. 13 and 74ff. 

16 On the old alphabet foimd in the Bodh Gaya monastery, S. Konow, AO 19, 
303f. Cf.B.Cn. Chhabea, Bulletin National Museum, New Delhi, no. 2 (1970), p. 

1'? H. MASPfeo, BEFEO 20 (pt. 2), 1-124; A. von Kosthorn, SAWW 219, no. 4; 
H, Jensen, Sign, Symbol, and Script, p. 179; R.H. van Guuk:, Siddham (Nagpur, 
1956), p. 36-44; K. Ch’en, Buddhism in China (Princeton, 1964), p. 478f. and 548. 
A natural application of this technique was in the transcription of Buddhist names 
of Indian origin. 

16 H. Jensen, Sign, Symbol, and Script, p. 211-214. 

19 H. Jensen, Si^, Symbol and Script, p. 198f. On the influence of Sanskrit 
upon phonetic studies in Chinese and Japanese cf. Kazuo Mabughi, Nihon inga- 
kushi no kenkyu, I (Tokyo, 1962). 


H. Scharfe * Grai3amatical Literature 

is almost certain: there is no other explanation for his adoption of a phonetic 
alphabet.20 The Indian influence on modern European phonetics is well docu- 
mented; in the words of J.R. Eirth: ‘‘Without the Indian grammarians and 
phoneticians ... it is difficult to imagine our nineteenth century school of 


The beginning of grammar was linked to god Indra. We read in Tait. Sanih. 
VI 4, 7, 3: 

130^ vai pardcy avyakrtavadat ; te devd Indram abruvann imdrn no vdcarn vydhurv 
iti . . . tdm Indro madhyato ’vakramya vydJcarot . . . ‘'Speech spoke turned away, 
inarticulate. The gods said to Indra: 'Articulate this speech for us!’ . . . Indra 
entered in the midst of it and articulated it,” 

Indra’s involvement with grammar is later attested in the Mahabhasya (I 5, 
25 f.: pupil of Brhaspati) while in the classical period it is always Siva who 
bestows grammatical wisdom (see below p. 92 fn. 24). 

Grammar was not a pure science; it was linked with the ritual duties of the 
priests who developed it. An example is the discovery of the cases in noun in- 
flection and how it was used to make the ritual more sophisticated. In the cere- 
mony of the repeated kindling of the fire (punarddheya or punarddhdna ) a cake 
is offered to Agm with six verses taken from the Rgveda. A condition is imposed 
that each verse must contain the name Agni in a different case form, as the 
varied incantation is more powerful. The termfor ‘case/ vihhakti]it, ‘distinction/ 
has been retained in later grammatical literature; but while it originally be- 
longed to the TQot'^hliaj ‘divide, share’ (Ait. Br. 1. 1 ; VTI 1), later authors some- 
times derived it from y^hhanj ‘bend’ (Katantra-vrtti 11 1, 2; III 1, 1,; cf. our 

Even though the verb was not yet recognized as a class, the categories ‘future’ 
{karisyat; Ait. Br. IV 29, 3, etc.), ‘present’ (kurvat; Ait. Br. IV 31, 3) and ‘past’ 
{kria; Ait. Br. V 1, 3)22 -^ere applied to verbal forms. This allowed for no distinc- 
tion between the different forms of the past (imperfect, aorist, perfect); it 
grouped the imperative with the future tense — ^in short, there was no attempt 
to describe the complicated structure of the Sanskrit verb. But the three time 
categories^ were there to stay — ^under different names. The later Brahmana-s 
and the Sahkhayana Iranyaka have hhavisyat, hhavat and bhuta instead.23 
With one more replacement, Panini has the definite set of hhavisyat, vartamdna 
and bhuta. 

20 g. Wild, ZDMG 112,294-297. That Sanskrit grammar was known in Sassani- 
dian Iran IS P^ven by the reference to Indian works on ^Py'^hrn (i.e. vyakarana 
grammar ) in Denkart IV 99-100; P. de Men-asce, JA 237 1-3 
J.R. FntTH, TPS 1946.119. 

^ KauMt^ Br^a^a XXH 3 has cakruas instead, possibly because the part. 

be interpreted as the name of the root kr: G.B. Paistob, JUP 
““f lately that Kau^itaki BrahmanaXXn 3 replaced krta with cakrvas 

because thus all three terms became active participles. ^ .van 

“ Mahabhasya I 254,13-166, 27 on whether roots denote 

The Origins 


It became clear at this time that many words belonged together as members 
of a word family, even if the exact nature of their relation was yet unknown. 
Due to the absence of a derivational hierarchy we find a motley group of head 
words. In the Ait. Br. the part. perf. pass, leads with about 50%, with agent 
nouns, action nouns, an adjective, an imperative, a third pers. sing, and a naked 
root24 sharing the rest. These forms are used with a suffix -vat having [. . .]' 
which derives from a priestly praxis to refer to a verse that contaius e. g. the 
word pra as pra-vat having pro" (Ait. Br. 1 10, 1). jdta-vat is thus the head word 
for a wide variety of words derived from the root y^jan, and rmhad-vat for words 
like mahdm, mahas, etc. In the Kaus. Br. the part. perf. pass, covers 90% of the 

The final redaction of the Vedic Samhita-s^s reveals the infiuence of grammat- 
ical thought in the mistaken extension of hiatus after the pronominal forms 
asme ‘us’ and yusme 'you’ in analogy to the dual forms ending in. -e, and in the 
compromise sandhi form [devdsjo a[pturah] Bgveda I 3, 8, etc. Because the 
Brahmana-s sometimes still quote the Bgveda words in their origmai pre-redac- 
tion form it can be assumed that the redactors’ work extended into the early 
Brahmana period ;26 for aU later authors the forms given in the sacred Samhita-s 
are beyond dispute. 

Major achievements were the Pada-patha-s ‘word-for-word recitations’ to 
the Vedic Samhita-s, probably iu the sequence Sama-veda, Rg-veda, Atharva- 
veda, Yajur-veda (first referred to possibly Ait. Br. V 4, 3 and certainly Ait. Ar. 
in 2, 6).^'^ The Pada-patha of the Rgveda is ascribed to a Vedamitra Sakalya. 
The whole text is broken up into its elements: sandhi between words is dis- 
solved, compounds are split up by inserting a pause between their members. 
Even certain case suffixes are separated from the stem, i. e. those suffixes before 
which the stem undergoes the same morphophonemic changes as between 
words: vacohhir (Bgveda 1 187, 11 ; instr. pi. of vacas) appears as vacahj Ibhih as 
against the instr. sg. vacasd. A late echo of this fjractice is foimd in Panini’s 
grammar where the noun stem before these suffixes is called pada ‘word.’ 
Occasionally the Pada-patha replaces extinct forms with those more familiar, 
e.g. usdsam (Bgveda IV 3, 11) with usasam and etana (Rgveda V 61, 4) with 
itaTia. The Pada-patha-s never achieved a status similar to the Samhita-patha; 

24 mad-vat (Aitareya Brahmana m 29, 2, etc.) may have been taken from the 
compound soma-mad ; in any case it comes close to the notion of the root. 

25 This is really a short form for samhitd-pdtha ‘recitation in connected speech’ 
contrasting it with the later Pada-patha (below). From sentences like pada-prakrtih 
samhitd “The samhitd [recitation] is made up from the word [-for-word recitation]” 
(Nirukta I 17 and Rgveda-prati^akhya II 1) samhitd could be taken to denote the 
text itself. 

26 L. Benou, Introduction g^nerale in: J. Wackebnagel, Altindische Gram- 
matik, I, 2nd ed. (Gottingen, 1957), p. 3, with literature. 

27 L. Benotj, Introduction generale, p. 118, fn. 571. A summary of their concepts 
is given by K. V. Abhyanxab, ABOR.I 54. 9-44. 


H. Scharfe • Grammatical Literature 

scholars occasionally disagreed with their analysis and did not hesitate to say 

Towards the end of the Vedic period there were thus three branches of linguis- 
tic study: phonetics (slksd), etymology (niruhta) and grammar (vydkarana)] 
but their oldest systematical works have not survived the hazards of oral tradi- 
tion. The relative chronology of the preserved texts allows no hiference on the 
genesis of the science. Inner logic as well as indirect testimony suggest that 
phonetics were the basis for the other two branches; actually phonetics and 
phonology are taken for granted by all authorities of etymology and grammar. 
This is especially clear in the sound table of Panini's grammar which is based 
on the grouping worked out by the phoneticians. 

The list of phonemes was more or less settled at an early time and what 
debates have survived deal only with such minor points as the status of visarga, 
anusvdra and other aUophones^® and with the question if /!/ needs to be listed 
or can be accounted for as a rare substitute for /r/.^o The sound table is generally 
phonemic and its authors deserve high praise for that; but it should be noted 
that at least one allophone (/h/) has shpped into the phonemic flock. Besides the 
vowels, semivowels and sibilants, the centre piece of the alphabet is the 
so-caUed pinca 'pa^m-vargdh [Rgveda-prati^akhya I 2(8)], the ‘five groups of 

Voiceless Stops Voiced Stops Nasals 

1. tenuis 

2. aspirata 

3. media 

4. media aspir. 

5. nasal 






























The places of articulation ( sthdna) move forward from throat (guttural) to lips 
(labial) ; the efforts of voicing and aspiration as well as the absence of them are 
well grouped. It seems that the beauty of the scheme led to the inclusion of /n/ 
wHch is reaUy nothing but a predictable allophone in Sanskrit (of. the corre- 
•spondence loc. sg. namani/ndmni vs. rajanijrajm).^^ 

^ L. Kenou, Introduction g^ndrale, p. 37. 

29 Katyayana, varttika 6 and 7 on Sivasutra 5, proposes to include these sounds: 
his su^estion is adopted in the larger recension of the Jainendra grammar, the 
bakapayana grammar, and other grammatical works. 

the phonetical description of /r/ (retroflex or dental) may reflect 
the historical fate of /r/ m its relation to /!/. 

M.B.Em^au, Lg 22.86-93. /n/ lacks contrastive value (there is no guna 

The Origins 


Every grammatical description presupposes some primitive etymology, if 
only to recognize that gacchati ‘goes/ jagdma ‘went’ and gatih ‘gait’ have a 
phoneme group and a notion in common. Etymology taken in this limited role 
may with Yaska ‘neglect the formation’ {na samskdram adriyda, Nirukta n 1). 
But etymology was not seen merely as a preliminary stage of grammatical 
analysis to be dropped later. It was as Yaska said ‘the complement of grammar’ 
(vyakaraimsya hdrtsnyam, Nirukta 1 15) ; a student is warned not to discuss his 
science with somebody who is ignorant of grammar (nirbruydt . . . ndvaiydka- 
randya, Nirukta II 3). We have to ask what this special contribution of nirukta 
was and why this branch of linguistics later became atrophic. Both questions 
are closely related. 

The Sanskrit word for grammar^^ is vydkarana lit. ‘separation, distinction’ 
(cf. Sat. Br. XIII 8, 2, 2 daivam caiva tat pitryam ca vydkaroti “He keeps separate 
what refers to the gods and what refers to the fathers”).^^ Grammar distin- 
guishes roots, suffixes, and prefixes and assigns each of the latter to a meaning 
or function. The interest is centered on forming correct words and sentences 
from these basic elements so that the intended meaning is expressed. The 
nirukta proceeds in the opposite direction: it explains words, especially Vedic 
words, by tracing the root or verb they are derived from.^ This process is 
called nirvacana ‘explicit mention [of the root/verb]’: in the Brahmana-s a 
verse was called a-nirukta if the praised god is not named in it.^s There is no 
interest in the formation of the word as long as it appears to be common and 
even if it is not, the etymology can proceed on the basis of the meaning alone 
with only the slightest formal support: 

32 Katyayana, in varttika 14 of his introduction, defines grammar as the sum of 
‘characterized [words]’ and ‘characterizing [rules]’: lak§ya-lak^ane vydkaraTiam. For 
Patahjaii (Mahabha^ya I 1, 19) and Bhartrhari (Vakyapadiya 111) grammar is the 
foremost of the six auxiliary Vedic sciences {veddnga lit. ‘limb of the Veda’); the 
others are phonetics, metrics, etymology, astronomy and ceremonial. One has to 
remember that in the Indian "way of speaking man has six limbs: arms, legs, head 
and sex organ. The Mimamsa disputed the claim of grammar to be a veddnga as e. g. 
Kumarila, Tantravarttika on I 6.3, 18-27 regards it only as smrti; cf. also K.C. 
Chattebji, JDLCU 24 (pt. 3), 1-21, summarized in J.F. Staax, A Reader, p. 289. 

33 P. Thteme explains vydkararia as ‘[word-] formation’: vividhena prakdrendkrta- 
yah kriyante yena; the sacred language is thus ‘built up’ or ‘prepared’ (samskrta): 
Studies . . . Fs. J. Whatmough, (’s-Gravenhage, 1957), p. 267-269. G.B. PAiiSTiLE, 
JUP 29.26-29, contrasts vyd-kr ‘to divide, analyse’ and sama-kr ‘to join, to synthe- 
size, to form;’ but he renders vdcain vydkaroti (Aitareya Brahmana 5. 1) with ‘utters 
speech’ (p. 27, fh. 38). Cf. above p. 80. 

34 ndmdny dkhydta-jdni “Nouns are derived from verb /roots,” Nirukta I 12 in a 

33 H. Oldeitbebg, Vorwissenschaftliche Wissenschafb — ^Die Weltanschauung der 
Brahmana Texte (Gottingen, 1919), p, 80; L. Renoit and L. Silbuun, Sarupa 
Bharati [Fs. Lakshjojst Sabot] (Hoshiarpur, 1954) p. 68-79. Even a grammarian 
like Patahjaii occasionally availed himself of the nirukta technique, e. g. Mahabha§ya 
I 206, 24 svayain rdjante: svardh “They shine by themselves: therefore they are 
called vowels ( avara ^ 


H. Scharfe * Grammatical Literature 

artha-nityah parlksetakenacid vrtti-samdnyena. avidyamane sdmdnye ^pyahsara-var- 
na-8dmdnydn nirhruydty na tv eva na nirbruydt, na samskdram ddriyeta, viiayavatyo 
vriiayo hlmvanW^ . . . examine them with regard to their meaning by the analogy 
of some [common] course of action. If there be no [such] analogy one should 
explain them even by the community of a [single] syllable or sound, but one 
should never give up [the attempt at] derivation. One should not attach impor- 
tance to the grammatical formation, for these complex formations are subject 
to exceptions” (Ninikta II 1). 

Looking at the noun forms kartre 'doer’ (dat. sg.), karma ‘deed’ (nom. sg.) and 
krtim ‘doing’ (acc. sg.) the etymologist would be satisfied to ‘explicitly mention’ 
the verb karoti ‘does’ in order to settle the basic meaning of the nouns ; the task 
of the grammarian would just have begun — but his interest in the notion of 
‘doing’ would be small. This becomes clear in the lists of motivations we find 
in two prominent works of the two schools. According to Yaska one should 
study nirukta in order to grasp the precise meaning of the Vedic stanzas; 
whoever does not understand the meaning cannot conduct a thorough investi- 
gation of accent and grammatical form, nor can he understand the word divi- 
sions of the Pada-patha; the attributes of the Vedic gods (e.g. archaic tvisita 
‘shining’) reveal their meaning only to the etymologist (Nirukta I 15 and 17). 
The grammarian is motivated in his study by his desire to guard the precise 
form of the Veda; to adapt the Vedic formulae to the people he is serving as a 
priest; because grammar is the most concise summary of the language; because 
it decides doubtful interpretations ; because it is his duty (Mahabhasya 1 1 , 14-2, 
2).36 We understand now why the oldest grammar we have, Panini’s Astadhyayi, 
avoids references to meaning as far as possible and why Panini’s root list 
(dhdiu-pdtha) originally was without lexical meanings. But there was a change 
later; sometime between Patanjah and Candragomin meanings were attached 
to the roots and thus grammars became complete language manuals. Etymology 
was retained in the schools of Veda interpretation, but there were no recasts 
of Yaska s classic work, let alone new nirukta-s for the Prakrits or other Indian 

What little we know about the period just preceding the great treatises comes 
from casual remarks in later texts, especiaUy Yaska’s Nirukta. The most im- 
portant discovery was no doubt the ‘root.’ It did not come easy or in one step. 
The Nighantu, an old list of rare Vedic words, gave nouns and verbs separate; 
particles and adverbs also were grouped together. The idea of a root was hazy 
at best: several times we find even different verb forms of the same root listed 
side by side. The most common form of quoting verbs was the 3rd sg. (jvalati, etc. ) 
which was found first in the Ait. Br. (V 20, 4, 8) and which has stayed on as an 
alternative mode of reference in all later grammatical literature. 

The imclear conception is reflected in an equally unclear terminology : dkhydta 
lit. ‘told’ is commonly used for the verb, while dhdtu lit. ‘part, element’^? is 

A second h^, Mahabhasya I 2,3 - 5,4, likewise stresses the formal aspects; 
one out of thirteen points refers to content 
L. Rbnou, JA 233. 138f. 

The Origins 


sometimes found in the same sense, but more often in the sense of the abstract 
entity called ‘root/ A third use ‘present tense stem’ can be inferred from the 
Paninian terms sdrvadhdtuJca ‘suffix attached to the whole stem’ (e.g. [bhava-jti) 
and drdhadhdtuJca ‘suffix attached to the halfstem = root (e.g. la-bhu-]t) before 
which a piece of the stem bJiava- is as it were cut off to get bhu-. An indication 
of how complex the problem is, is in Yaska’s statement: 

tatra ndmdny dhhydia-jdniti Sdkatdyano nairukta-aamayai ca “Amongst them the 
nouns are derived from the dkhydta; so says Sakatayana and the doctrine of the 
etymologists” (Ninikta 1 12). 

Did Sakatayana derive the nouns from verbs? Or did he use dkhydta for 
‘root’? If he did, the anonymous author quoted in the Mahabhasya (II 138, 14 
and 16) was right when he said in his own terminology: 

nama ca dhdtu-jam aha Nirukte, vydkarane Sakapisya ca tokam “The noun is 
derived from the root; so he said in the Nirukta and [so said] in grammar the 
offspring of Sakata.” 

Sakatayana who evidently had claimed that all nouns are derived from 
dkhydta-B, is opposed by others: na sarvdniti Qdrgyo vaiyakaraTmndin caike “Not 
aU — ^thus [say] Gargya and some of the grammarians” (Nirukta I 12). Prom 
these sentences it is clear that Sakatayana was not an etymologist but a gram- 
marian and that Gargya was not a grammarian — ^but Gargya cannot have 
been an etymologist either because the etymologists subscribed all to the doc- 
trine of general derivability.^s Since a Gargya is credited by tradition with the 
creation of the Pada-patha of the Samaveda, the conclusion is plausible that 
the Gargya mentioned by Yaska was a follower of the siksd. Eds influence on 
some grammarians was of great consequence because Panini followed him rather 
than Sakatayana (cf. below p, 104). 

Another famous controversy goes back to this period: do prefixes have a 
meaning of their own (so Gargya)^® or do they merely illustrate (dyotaka) a 
meaning contained in the root (so Sakatayana) ? This debate, reported in Nirukta 
I 3, has later continued in the grammatical literature.^ Panini again seems to 

28 According to an old saying, he would have been the founding father of gram- 
mar: anu Sdkatdyanam mi 2 /dfcararid^“ All grammarians are followers of Sakatayana” 
(Ka^ika on Paruni I 4 86) ; cf. Yudhisther Mimamsak, Samskrt vyakaran ka itihas, 
I (1963), p. 162f. Patanjali (Mahabhasya II i20, 20f.j has an anecdote that 
Sakatayana was so immersed in thought that he did not notice a passing caravan 
of carts (sakapx) ; the play on his name may well have been the cause of the story. 
In a verse quoted in Mahabhasya II 138. 16 he is called the ‘offspring of Sakata.’ 
Sakatayana admitted only one root per noun (i.e. root -h suffixes) while others used 
up to three roots to explain a single word, e. g. hrdayam ‘heart’ from ^hr -|-]/dd -^-Yyam 
(Sat. Br. XIV 8, 4, 1). 

29 L. Renotj, Introduction gen^rale, p. 39. 

^2 If Gargya was the author of the SV Pada-patha it fits perfectly that in this 
work prefixes are separated from the rest of the word: e. g. aam itdram for aamvdram. 

D.S. Rxtegg, Contributions al’histoire de la philosophie linguistique indienne 
(Paris, 1959), p. 25, and K.V. Abkvaitxab, A dictionary of Sanskrit grammar 
(Baroda, 1961), p. 81. 


H. Scharfe • Grammatical Literature 

follow Gargya because be specifies in 1 4 93 adhi-parl anarthakau ''adhi and 
pari when they are meaningless this restriction is only appropriate if they 
usually are meaningful. 

Grammatical analysis in itself was challenged by thinkers like Audum- 
barayana for whom the sentence was the linguistic unit ; the reference to words 
in grammar is valid only because they allow a complete and convenient way to 
handle the language.^ 

Although Panini and Yaska mention a good number of previous scholars on 
grammar, etymology and phonetics (^akatayana and Gargya are mentioned 
by both^), we know very little about them.^ Works running under their name 
are spurious^s and alleged quotations in the later commentatorial Kterature are 
as a rule highly suspect.^s Their insights were assimilated by their followers, 
but their compositions were lost when the classical works of Panini and Yaska 
rose above the previous literature.^'^ Oral tradition knows no mercy for outdated 

The scholarly literature of this period developed a new style which calls for 
some comments.48 The new texts called sutra-^ ‘threads’ aimed at extreme pre- 
cision and brevity; their single sentences, likewise called sutra-s, reduced or 
eliminated finite verb forms and dependent clauses. Often they attained further 
economy by dittoing expressions that normally would be repeated in several 

^2 The best explanation of the controversial paragraph in ISTirukta 1 1-2 has been 
given by J. Beough, BSOAS 14. 73-77. 

The references to Gargya and Sakatayana in the Prati^aldiya-s have been col- 
lected by A. C. BuEimiJi in his edition of the Rktantra-vyakarana (Mangalore, 

^ The memory of a few rules of Api^ali seems to have survived: F. Kxelhoen’, 
lA 16.102. 

^5 The Sakatayana-vyakarana we have is a Jain grammar dating from the tenth 
century A.D. 

Whenever Patanjali, Kaiya^a, etc. encounter an imusual term or expression 
they offer readily the answer that this is a holdover from the puTvdcdryci-B ‘earlier 
teachers;’ but they never give specific references which they probably would have 
had such been available to them (F. Kjelhoen, IA 16. 101 f.). Such explanations 
should be viewed like other suggestions pointing out a possible cause for the appar- 
ent anomaly, not a statement of fact, cehrlyita ‘intensive middle’ (MahSbha^ya U 
232, 5 in a quote) is explained by Kaiyata as a term of the puTvdedrya-s but in Ni- 
rukta VI 22 carkarlki (which denoted an active intensive in II 28) comprises the 
middle as well: this makes cehrlyita a later fabrication (L. Rekou, JA 233. 166 fn. 2). 
For a possible pre-Paninian au^ (in Pan. VH 1 18) *nom./acc. dual -aw’ pleads 
M.B. BAIiAStmEAHMAKYAM, JUP 25.77-82. 

Pauim quotes ten authors besides the general dcdrydndm ‘according to the 
teachers’ (which teachers?). Katyayana v^tika-s 16 and 17 on I 1 44 notes that 
if words were not permanent one would have to conclude that forms quoted for a 
certain teacher or area are valid in that sphere only. But the prevailing dogma of 
aabda-nityaiva ‘permanence of speech sounds’ must accept all these forms as equaUy 
correct. For the later commentators (Kaiika, etc.) such rules are just optional and 
the use of names (instead of the shorter vd ‘or’) serves to honour earlier scholars. 

L. Rekotj, JA 233. 106-165 and JA 251. 165-216. 

The Origins 


consecutive sentences (anuvrtti). This made it easier to memorize whole text- 
books but made interpretation more problematic. The memorized sutra was 
accompanied by the teacher’s interpretation that ‘fleshed out’ the skeleton 
information suppKed in the svira. This puts us at a disadvantage as we generally 
lack such an authentic interpretation. Because the teacher’s interpretation was 
not as rigidly recorded as the sviira-B and because the wording of the sutra-s is 
often ambiguous, inner developments in a school could creep into the traditional 
interpretation almost unnoticed ; sometimes different branches of a school gave 
different interpretations of the same siitra. 

The name for this style is taken from the image of weaving where a thread 
is stretched out lengthwise as a w^arp to be crossed by the woof.'^® The warp may 
be one continuing thread or it may be cut on both sides of the frame: this 
explains the use of sutra for both the whole work and its sentences. The sutra is 
thus a stripped textus. This explanation is supported by the parallel case of 
tantra ‘thread, text’ with its counterpart dvdpa ‘insertion.’ In liturgical texts 
one discerns a) the special feature of a ritual (pradhdna) and b) the auxiliary 
rites that accompany various rituals in an identical manner (anga): from a 
strictly formal point of view the ancient ritualists could call these auxiliary rites 
which turn up again and again, the warp ( tantra) and the main rite which va- 
ries from ritual to ritual, the woof ( dvdpa ) 

49 P. Deussen, Das System des Vedanta, 2nd ed. (Leipzig, 1906), p. 27. The 
Tamil grammar Nannul (verses 24 f.) offers two interpretations: a) spinning yam 
(the text) from cotton (wool) (the words) and b) the use of a carpenter’s line to draw 
markings on imeven wood. Could one connect the image of spinning with the expres- 
sion vHti ‘turning’ used for the teacher’s elaboration in class and anuvrtti for dittoing ? 
Others have suggested that the underlying image is that of a necklace held together 
by a thread ; this explanation leaves the double use of sutra for ‘text* and ‘sentence’ 
xmexplained. W. Buben, Einfuhrung in die Indienkunde (Berlin, 1954), p. 146f., 
suggested a thread ‘script’ comparable to the Inca quippu; Th. GoLDSTijcKEB, 
Panini (London, 1861 ; reprint Osnabriick, 1966), p. 24r-26, saw the reason for the ab- 
breviated style in a desire to save writing paper and identified sutra with the thread 
that held the manuscript leaves together. The expression ‘fils conducteur’ pli. 
Benotj, Terminologie grammaticale (Paris, 1957), p. 341] is based on the Greek 
legend of Theseus and Ariadne in the labyrinth and hence not applicable here. 

50 L. Benoit, JA 233. 113, fh. 1 and Vocabulaire du rituel vedique (Paris, 1954), 
p. 30 and 69f., and G. Abtoua, WZKAI 52. 382-385. tantra and dvdpa occur together 
also in &abarasvamin’s Bha$ya on Mimaipsa-sutra XI 1,1. 

Chapter II 


Panim is not only the oldest grammarian whose work has come down to 
ns — he is also the greatest. His full name was Daksipntra Panird (Mahabhasya 
1 75, 13 and III 251, 12 in a quoted verse) in the common style which added a 
metronymic to the name itself. Even though in the old times the father’s caste 
alone decided the caste of a son (cf. the story in Chand. Up. IV 4), it could only 
increase a Brahmin’s prestige if he could point out the equally high caste of his 
mother.^ Panini’s date can be fixed only approximately; he must be older than 
Katyayana (c. 250 B.C.) who in his comments on Panini’s work refers to other 
stni earlier scholars dealing with Pamni’s grammar; his proximity to the Vedic 
language as found in the Upanisad-s and Vedic sutra-s^ suggests the 5^ or 
maybe 6*^ c. B. C.® His home was the town of Salatura^ in ancient Gandhara 
(northwest India), modem Lahur in the angle where the Kabul Hiver and the 
Indus meet, about 20 miles northwest of strategic Attock Bridge. Outside 
Lahur, there are now some high mounds which likely contain ancient sites. The 
Chinese monk Hsuan-tsang visited the city in the 7^ c. A. D. ‘‘To the north-west 
of TJ-to-Ma-han-c’ha (i. e. Attock), 20 li or so, we come to the town of So-lo-tu-lo 
(Salatura). This is the place where the Ksi Panini, who composed the Ching- 
ming-lun (vyakataim), was bom. ”5 retold also a Buddhist legend in which a 

1 P.V.Kane, IHQ 14.239f.; P.Hoesch, As. St. 18/19.227-246; T.R. Traut- 
MANN, JPAS 1972. 12f. 

® F. Kielhoek, GGN 1885.186f.; B. Liebich, BB 10.205-234; 11.273-315 and 
m his book, Panini (Leipzig, 1891), p. 38-50; 0. Wecker, BB 30. 1-61+177-207; 
P. Thteme, Panini and the Veda (Allahabad, 1935), p. 75-81. 

^ There ^ no shortage of inconclusive arguments. V.S. Aoeawaxa, India as 
toown to Panini (Lucknow, 1953), p, 380-383, assumes that the word masJcarin 
carr^^ a bamboo cane > religious mendicant’ in VI 1 154 refers to Makkhali 
Gos^putta the older contemporary of the Buddha and advocate of determinism, 
^ ® reference to a certain person. If P. H. L. Egoeemoitt, Persica 

4.88-97 is nght m fixing the date of Buddha’s death at 384 B. C. instead of 484 B. 0. 

the mm of some Vedic texts (and along with them Panini) could come down a 
nunared years. 

Salaturlya ‘man from Salatura’ in an inscription of giladitya 

Wa W V P- 176, mBhamaha’sKavy^lam- 

kara yi 62 md m VMdhamana’s Ganaratnamahodadhi, commentary on verse 2. ' 

fvt DO Western World’ by Hsiian-tsang, trans. by S Beal 



Brahmin claims that there was even a statue of Panini in existence. Panini’s 
origin in the extreme Northwest explains several pecuharities of his grammar. 
He describes in minute detail features of the area as e. g. the different accents in 
the names of wells north and south of the river Vipas (IV 2 73, 74 cf. below p. 108), 
while he limits himself to the larger geographical features of other parts of 
India. Furthermore, he shows great familiarity with the Vedic texts found in 
the North and West (notably certain branches of the Black Yajurveda), but 
never quotes forms from the White Yajurveda belonging to the East.® As the 
province Gandhara was an Achaemenian satrapy from the time of Darius 
to the end of his dynasty, Panini must have been a Persian subject at least in 
name,’ but his work shows no trace of it.® 

Panini’s grammar, the Astadhyayi ‘The Eight Chapters,’ is composed in a 
most concise sutra style and consists of a Httle under 4000 sutra-s. The exact 
number varies depending on our judgement of a few interpolations and ‘sutra- 
splittings.’ Most of the interpolations were originally varttika-s of Katyayana 
or are at least based on Katyayana’s or Patanjali’s remarks; more frequently 
only the wording of a sutra has been altered or enlarged on the basis of such 
remarks. As only the text in continuous recitation ( samhita-paiha) is authorita- 
tive, the Paniniya-s often resorted to dividing a sutra into two or even three: 
it allowed them to reinterpret the rules without actually altering them.^ 

Our vulgate text is the one explained in the Kasika-vrtti (7^ c. A.D.); it is 
about 20 sutra-s longer than the text known to Katyayana or Patanjali.^® The 
fate of the text before Katyayana is not known. It seems that the oral tradition 
suffered some interruption, for the technical accents and nasalizations which 
are an essential element of Panini’s metalanguage were lost and already Katya- 
yana probably had to hifer their existence from the context or from the results, 

6 P. Thteme, Panini and the Veda (Allahabad, 1935), p. 74-79. 

’ J. Filliozat, JA 240.321. 

® Of no historical value are the legends about Panini in later literature such as the 
Pahcatantra (ed. G. Buhleb) II 33, Kathasaritsagara I 4, 20-25 or Haracaritacin- 
tamani XXVH [ed. & tr. J. Brough, Selections from classical Sanskrit Literature 
(London, 1951), p. 2-21] ; some poetical works attributed to him are really products 
of the late Middle Ages (F.Kielhobh, GGN 1885. 185f.). 

9 F. KrEiiHOBH, lA 15. 203 counts 3983 sutra-s (cf. the numbering in Bohtlingk’s 
edition); L. Rehou, La Durghatavrtti (Paris, 1940), vol. I, 1, p. 10, totals 3996 
sutra-s. It was of course impossible to recite a whole adhydya or even a pdda without 
interruption. How do we account then for the assumed continuous recitation? Did 
the Astadhyayi go through a stage of written tradition in which accents and nasali- 
zations were lost and all rules subjected to sandhi (P.T h i em e, Paruni and the Veda, 
p. 120-124)? 

19 F. KIielhobh, IA 16.178—184. The enigmatic sutra-s I 2 53-57 with their 
argumentative style must be an interpolation, and their tenets point to a different 
school of thought. As only the first of them is commented on and mentioned in the 
Mahabhasya (Patahjali only), it is likely that the others got into the text later; 
Pataiijali could otherwise hardly have avoided any comment on them [B. Fadde- 
GOH, Studies on Panini’s grammar (Amsterdam, 1936), p. 57-59; L. Renou, JA 
233. 115 fn. 3; G.B. Palsule, ABORI 30. 135-144]. 


H. Scharfe ' Grairrmatical Literature 

i, e. the known forms of the Sanskrit language. In his remarks on 1 4 1 Patanjali 
mentions a valiant reading: 141 was read both as a kaddrad ekd samjnd and as 
prdkhaMrdtparamkdryam (Mahahhasya 1 296, llf.).^^ Besides the 1713 sutra-s 
discussed in the Mahahhasya, many others are incidentally mentioned in it, 
thus providing testimony of their existence. We have no reason to doubt the 
basic validity of our Panini text ; its limitations are in the phonetic presenta- 

The loss of the technical accents^^ ^nd nasalizations^^ was already mentioned. 
Other changes may have been caused by written fixation of the text. The 
technical language had— due to a combination of determinatives with suffixes 
of the Sanskrit language — a number of unusual consonant clusters: Hvd, 
etc. In many instances, the clusters were so unfamiliar to the Sanskrit speaker 
that he heard a hint of a vowel: hTHa^, tasd^ (but etc. 

Whether Paijini consciously used an auxiliary vowel to facilitate pronxmciation 
(uccdrandrtham)^^ or whether it originated in the mind of the listener we do 
not know.i7 If the latter was the case, a serious problem must have arisen when 
the grammar was put down in writing: we find the three short vow'^els (a, i, 
used with no apparent reason for their distribution except that the same vowel 
is used for a certain suffix in all its occurrences (HTHa^, In one case 

the new vocalization led to a secondary differentiation: the suffix 3 ^ was ren- 
dered as 8u<^ when attached to numerals (e.g. dvi-s) but as when used as an 
aorist marker (akdr-s-it) even though it had the same formal characteristics.^® 

F.Kielhoen, Gurupujakamnudi [Fs. A. Weber], p. 32. Many Vedic texts now 
without accent marks have suffered a similar loss. The loss of technical nasalization 
meant that the abbreviation for all case suffixes from a* to sup (I 4 14) became 
a homonym of the loc. pi. suffix auP. 

-7 m 2 134 a ’‘VEs tacchUa-taddharma-tataOdhukarifu is also read 

. (Mahabh^yall 135, 19f.). Occasionally Patanjali and repeatedly the 
flif /V varitot readings or alternative interpretations to Panini him- 

selt (V.S. AgbawaIiA, India as known to Panini, p. 30f.). 

A?tadhyayi mentioned by Ytohisthib 
S anKkyt vyakaran ka itihas (Bahalgadh, 1973), vol. I, p. 230, must be 
recent attempts to reconstruct the original form of Panini’s work. 

pronuncation retained in the rule I 1 
but a LofTn f ® ^ is not a technical form of Panini’s 

Ld the vSa!p. m). Padapatha of the Rgveda (P. Thibme, Pknini 

w f suffixes found in kutas and kutra, respectively. 

the Bntish acronym WEENS for ‘Women’s Royal Naval Sewice’ with an 

IS^bv •’ the teaditional ^terprl^ltion 

supported by A. Wezxbb, Kratylos 18.24-26. 

is afSf '“1 ‘5t *•““»“<> aislribrtion of ths» vomte 

■' •” * 

H. ScHARFE, Panini’s metalanguage (Philadelphia, 1971), p. 7-9. 



Finally, Panini had a unique way to indicate the short, long or protracted 
duration of a vowel: he likened it to the crow of a cock: “The short, long and 
protracted [vowel] respectively have the length of u, u, There was no 

proper way to put down the last imitative sound in writing and the three vowels 
were contracted into a single long /u/: u-Jcalo ^j-jhrasva~dirgha-plutah (I 2 27); 
the com m entators from Katyayana on had to reach the true form by logical 

Panini’s grammar was clearly meant to be memorized before it could be 
studied as any operation or definition involves numerous leaps from chapter 
to chapter, forward and backward. And indeed it was the custom for students 
to learn the grammar by heart at a tender age before the teacher explained its 
application.^! To the uninitiated the first impression is one of organized chaos. 
This is largely due to a constant struggle between two principles: logical group- 
ing and technical economy. While the logic of things has settled the basic 
structure of the work, the quest for economy often prevailed in its execution. 
Besides, there are numerous associative digressions that disrupt the basic 
scheme. L. Eenou^s has pointed out that Katyayana’s proposals if inserted into 
the grammar at their points of reference would create much the same picture, 
and he has concluded that these digressions (or at least many of them) may be 
interpolations and indicate the gradual growth of the grammar. I would add 
that these interpolations are probably the work of Panini himself. The theory 
that Panini was only the last in a line of redactors of a traditional text^^ lacks 
proof and even probability because many of the ‘interpolations’ are indispens- 
able parts of any Sanskrit grammar. 

-0 Nagojibhatta’s explanation: “a, etc. are not given because the quantity of 1, 2 
and 3 measures is established for the -w-sound crowed by cocks” {kukJcuta-ruteukdre 
eka-dvi-tri-matratva-prasiddher akarddayo ndktdh Laghu^abdendxriekhara on I 2 27, 
KSS no. 128, p. 67) is based on Vitthala’s commentary of the Prakriyakaumudi I 2 
27, BSPS no. 78, p. 21: “Because the status of 1, 2 and 3 measures is established in 
the tt-sound in the cock’s crowing the author of the sutra used the ti-sound setting 
the a-sotmd aside.” Otherwise Indian phoneticians have compared the length of 
speech sounds to that of typical sounds of different animals: “A blue jay cries one 
measure, a crow two measures; a peacock should be recognized as having three 
measures” [Rgveda-prati^khya Xm 20(50)]. Of. B. Liebich, Einfiihrung in die 
indische einheimische Sprachwissenschaft II (Heidelberg, 1919), p, 42, and P. 
Thibme, Panini and the Veda, p. 125 f.; not convincing G. Cajidona, Lg. 41. 236 f. 
The pitches of the 7 musical notes are measured against animal sounds, from a 
peacock’s cry to the trumpeting of an elephant [Narada Siksa, §S, p. 407, quoted 
by S. Vabma, Critical studies in the phonetic observations of Indian grammarians 
(Oxford, 1929), p. 159]. 

2! This is the common practice down to modem times (J. BAiiLAOTYNE, in The 
Pandit 1.146-149). 

22 Lf. Renoit, Introduction generale in J. Wackemagel, Altindische Grammatik, 
I, 2nd ed. (Gottingen, 1957), p. 116 fn. 545. 

23 I.S. Pawatb, The stmcture of the Ashtadhyayi (Hubli, 1935??), p. 114, and 
R. Bmw6, Studien zu Adhyaya m der A^tadhyayi Paninis (Wiesbaden, 1966), p. 93 
and 147-150. 


H. Scharfe • Grammatical Literature 

The grammar proper is preceded by a sound table^^ based on the old phone^ 
mie alphabet worked out by the siksa. Panini reduced the vowels (a, a, i, I, etc.) 
to their types (a, i, etc.) comprising all varieties by duration, accent, nasaliza- 
tion, etc. The original sequence of sounds was modified in several ways; then 
the list was broken up into 14 sections, each with another consonant as end 
marker. This arrangement enabled the author to form groups almost at will by 
combining a sound of the list with an end marker further down: this combina- 
tion comprises all sounds down to (and excluding) the end marker, e.g. in the 
sutra-s 1 a t 2 r Z* the abbreviation a? would denote the vowels a, u, the 
abbreviation the vowels u, r, 1. This is done by a process of contraction 
called samdhdra 'gathering’ introduced in rule 1171, adir dntyma saMtd “The 
beginning with the determinative that is last.” Short labels thus denote whole 
classes: ha^ all consonants, ya^ all consonants except /h/, kha?f all voiceless stops, 

These formulae are used frequently in the grammar as in the very first sutra 
of the grammar proper: mddUr “[When I say] vrddhi [I mean] a, ai, au:' 
In this rule the usual word order has been inverted to obtain an auspicious 
beginning: vrddhi means ordinarily ‘growth, prosperity.’ This is taken up at the 
end of the grammar with the expression udaya which ordinarily means ‘success’ 
but here is used in the meaning of 'following, The very last sutra VIII 4 68 
a a reverts to the sound table at the beginning: it corrects an assumption made 
throughout the grammar-~that the short /a/ is phonetically homorganic with 
the long /a/. After we have used /a/ as just a short open a, we now learn that its 
phonetic value is that of a closed a. The phonemic oppositions contrast with the 
phonetic ones more in this way: 

Phonemes Sounds 

















thrift ‘contraction/summation rules'; under 

^iva.3iara.s and mahe^vara-rntra-B 
p 3^381 ^ ^ [K.M.K. Sabma, Pamni Katyayana and Patanjali (Delhi, 1968), 



The jSrst chapter is basically^^ taken tip by 'meta-rules/ rules about the rules. 
Some of them are definitions (sarrijna-sutra) ; others, rules of interpretation 
(paribhdsd’Sutra). In the definitions, the defined term (samjnd) stands last, 
the content (samjnin) fimt (cf. Mahabhasya I 40, 9), e.g. I 1 60 a-darianam 
lopah 'lopa [means] not-being-seen’ i.e. the word lopa lit. loss, deprivation’ 
denotes no such physical action but merely states that a certain sound or suffix 
will not be seen. In his meta-rules Panini established a special technical language 
(meta-language) to facilitate the grammatical description for which he uses the 
term upadesaMt. ‘instruction. The creation of a metalanguage meant in the first 
place new word-like elements the suffix of the part. perf. pass, -ta gets a deter- 
minative * to indicate the weak grade of the preceding element (e. g. Iup4a) ; a 
suffix -a may get the determinatives ^ and the ^ to indicate in the preceding 
stem the loss of the last vowel and any consonant that follows, and the V’ to 
indicate vrddhi for the first vowel of the stem (^aV' V 1 62 ; thus we form from 
trims lat ‘30’ trairm-a ‘containing 30 chapters’ as the name of a text). 

New values are assigned to several cases: the genitive is to be construed with 
a supplied sthdne ‘in place of . . .’ {^asihl sthdne-yogd 1 1 49),3o the locative deno- 
tes the following element (‘before . . .;’ I 1 66), the ablative the preceding one 
(‘after . . 1 1 67). Often several such meta-rules are involved in the formula- 
tion of a single sutra. In the manner described above the vowels i, u, r, I, short 
as well as long, can be summed up in the contraction likewise aU the vowels 
from atoaum the contraction and the semivowels y, v,r,lm ya^. If we add 
now the proper case endings as per meta-rules we get the sutra VI 1 77 
yaV' a^,i “Instead of ya^ is substituted before ; as both the substitutes 
(?/, r, V) and substituends (i, ti, r, 1) are equal in number, the substitution will 
proceed strictly parallel (cf. Panini’s rule I 3 10), corresponding to Whitney’s 
rule number 129: “The i- vowels, the u- vowels, and r, before a dissimilar vowel 
or a diphthong, are regularly converted each into its own corresponding semi- 

27 Other rules entered the chapter by association. 

28 A grammar of it is H. Schakfe, Panini’s metalanguage (Philadelphia, 1971). 
Because references in grammar are usually to word forms, references to content are 
marked with an iti — a direct reversal of the common Sanskrit practice: Katyayana’s 
varttika 3 on I 1 44 iti-karano ^rtha-nirdesdrthah is used to make it refer to 

29 The attached determinatives (it) disappear -when the Sanskrit word is formed, 
before the application of any rules. From that follows the explanation of their name: 
i-t ‘the one that goes away’ given by Candragomin in his vitti on 1 1 5, Abhayanan- 
din on Jainendravyakarana 12 3, commentary on Sarasvatikanthabharana 12 6 
(cf. the term krt ‘primary suffix’ from lit. ‘that which makes [a noun]’; the 
Nirukta I 17, II 2, etc. uses ndma-karana ‘noun maker’). Less convincing is L. 
Rbnotj, Altindische Grammatik I (2nd ed.), Introduction, p. 41, with fn. 615: short 
for iti. In his Terminologie Grammaticale du Sanskrit {2nd ed.), Renotj accepted 
the interpretation *ce qui s’en va.’ Both publications appeared in 1957. 

20 On this use of a postposition-like sthdne ‘instead of,’ see A. Wezleb, KZ 86. 
14^19; on the use of yoga cf. II 3 16, V 4 44 -f 47 + 50, etc. 


H. Scharfe • Grammatical Literature 

vowel, y or V or /!/ though, formally included in Panini’s rule has no applica- 


In the root list appended to the grammar, the dhdiu-pdtha, many roots were 
marked with one or two technical accents to subject them to certain rules of 
the grammar (below p. 101) ; most of these accents can be reconstructed exresul- 
tatu. Besides, the svarita-a^cGent was used to denote adhikdra-s ‘charges’ (1 3 11). 
The problem is that we no longer know for sure which of Panini’s rules were 
adhikdra-s and what their functions were. Headlines like III 1 1 pratyayah 
“Suffix” were easily recognized as such and hardly needed additional marking. 
Occasional remarks in the Mahabhasya have led to another interpretation, viz. 
that these ‘charges’ were the expressions that are to be dittoed in the following 
rules. A difficulty is that the end of the dittoing is not marked as well. There is 
frequent overlapping and boxing; it is by no means true that a ‘charge’ would 
end where the next begins. Another possibility is that a marked ‘charge’ could 
refer to a certain section in the grammar, e.g. the word strl- in I 2 48 if marked 
as a ‘charge’ could refer to the a^nyam-section IV 1 3-81.32 

Under the heading kdrake ‘for a factor [of the action]’ Panini introduces six ba- 
sic semantic notions (1 4 23-55) that play a crucial role in his syntax and indirectly 
also in his morphology. Their number and character make it likely that these 
notions were originally conceived in analogy to the cases (excluding the vocative 
and the elusive genitive) : 

14 24 dhruvam apdye ^pdddnam “[That which is] firm when depai'ture [takes 
place] is [called] apadana ‘take-off’ ” 

32 IcarmaTm yam ahhipraiti 8 a sampraddnam “He whom one aims at with the 
object is [called] sarnpraddna ‘recipient’” (lit. ‘bestowal’) 

42 sadhahatamam karanam “That which effects most is [called] karana 
‘instrument’ ” 

45 adharo ^dhikaranam “Substratum is [called] adhikarana ‘location’ ” 

49 kartur Ipsitatamam karma “What the agent seeks most to attain is [called] 
karman ‘deed, object’ ” 

54 8va-tantrah kartd “He who is independent is [called] hartr ‘agent’ ”33 

The objective relations and the kdraka-^ do not run completely parallel and 
therefore the hdraka defimtions are followed by amendments which substitute 
one kdraka for a n exp ected other. By the definition I 4 45 we expect the sub- 
stratum of adJiiysthd rule’ to be adhikarania ; but we are told in I 4 46 that it is 

31 W.D, Whitney, Sanskrit grammar (Leipzig, 1889), p. 44. 

32 P.EhELHOEN, GuTupujakaumudi [Fs. A. Webeb] (Leipzig, 1896), p. 29-33; 
essentially a re-statement (without reference to Kjelhobk) is G. Cardona, in 
Pratidanam [Fs. F.B. J. Kuipbb] (The Hague, 1968), p. 448-454. 

^sote that karman (defined in rule 49) is already used in the definition rule 32 ; 
iikewose kaHf of 54 in 49. But there is no circle: kartr, the last, is the kingpin. The 
se^enee m which the definitions are given is the consequence of the rules I 4 1-2 
as Katyayana yartt, 30-34 on 1 4 1 demonstrates: each item in this section can have 
DONA^^OTh^43^f of conflict the one taught later prevails (G.Cab- 



harman instead. Consequently T^e use (by II 3 2) the accusative grdmam 
adhitisthati C‘He rules in the village”) and not (by 113 36) the locative *grdme dhi~ 
tisthati ; and we form a passive grdmo dhisthlyate instead of "^grdme 'dkisthiyate. 

If the notion of kdraha-s was perhaps derived from an observation of Sanskrit 
cases, Panini has raised them above the level of case values and made them 
intermediaries between reality and the grammatical categories. Their impor- 
tance, often misunderstood, goes far beyond the s3Titax of cases; next to the 
roots, they are the prime moving factors of the whole grammar. There are, as 
Katyayana vartt. 5 on II 3 1 points out, five ways to express a kdraha relation: 
case suffix, verbal suffix, hrt suffix, taddhita suffix and compound. 

The use of cases is taken up only late in the next chapter (11 3 1-73), well 
separated from the kdraha-section. Though the kdraha values are the basis for 
the assignment of case suffixes, again there is no one-to-one relationship 
While basically the second case ( = accusative) suffix is used to denote the object 
(karman)^ in the Veda also the third case (= instrumental) suffix can be used 
(II 3 3-4). The number of such modifications in the case syntax section is larger 
than in the kdraha section which accounts for its greater length. The sequence 
in which the cases are treated is different from the morphological ranking of the 
cases as it appears in their names {pratkamd ‘first’ = nominative, dvitiyd 
‘second’ = accusative, etc. based on the suffix list IV 1 2) nor is it related to 
the sequence of the kdraha definitions. It seems that the sequence is based on the 
convenience of dittoing expressions from one sutra onto some following sutra-s. 

The section begins with the sutra II 3 1 anabhihite which is dittoed through 
n 3 71 (the end of the section) : “When it (i. e. the kdraha named in the respective 
rule) is not already expressed [otherwise].” Evidently, the ease syntax rules 
are conditioned by a prior application of the kdraha definitions. The second case 
(= accusative) suffixes are used to denote the object (harman)^ the third case 
( = instrumental) suffixes for the agent (kartr) or instrument (karana), etc., for 
sundry relations the sixth case ( = genitive) suffixes^® — all provided that these 
semantic relations have not already been expressed otherwise. The first case 

S'* Of course, the divergence would be even greater if we tried to pair the cases 
with the objective relations; cf. also G. Cabdona, JOIB 16.207-210. 

S5 The horizontal arrangement with the nominative first and the accusative 
second leaves the strong forms of the noun together; if read vertically (as in the 
jpunarddheya rite, cf. above p. 80) it allows maximal dittoing of forms. 

36 The rule IE 3 50 saspil sese “The sixth [case ending] for the rest” reflects the 
difficulty to define in a few words the use of the genitive case ; cf. A. Debbxjkneb, 
Aus der Krankheitsgeschichte des Genitivs (Bern, 1940), E. Schwyzeb, Griechische 
Granunatik, 2nd. ed. (Munchen, 1959), vol. II, p. 89; and also cf. Mahabhasya 1118, 
10 eka-satam sasthy-artdh “101 are the meanings of the genitive [suffix]” and I 463, 
13 karmddinam avivaksd sesak ‘‘sesa ‘rest’ is the non-intention to express object, 
etc.” Por the residual form of the statement, cf. I 4 51 akathitarn ca “Also that which 
has not been told so far [is called ‘object’].” Bhartrhari’s Vakyapadiya III 7, 156 
treats sesa like a kdraha, and Amaracandra’s Karakanirupana lists it as the seventh 
kdraha [Descriptive Catalogue of Sanskrit Manuscripts in the Adyar Library, VI 
(Madras, 1947), p. 151]. 


H. Scbajpfe * Grammatical Literature 

( = nominative) suffix is added when we want to express merely tte gender and 
nnmber of the nominal stem notion.^’ 

In 1 3 1 bhuv-ddayo dhdtavah^ Panini defines dhatu ‘root^ by referring to the 
root list (dhdtu-pdtha) appended to his grammar “Roots are ]/6^ etc. (as 
listed).” He takes up the subject in II 4 35-57 when he teaches root substitu- 
tions for defective verbs like hanivadh, asjbhu that supplement each other in 
noun formation and verbal paradigm. With the word nucleus thus established, 
the process of growth — adding suffix after suffix like a tapeworm growing from 
its head — ^begins with the sutra III 1 1 pratyayah “[What follows now is a] 
suffix.” The first suffixes taught are closest to the roots; in fact Panini declares 
that they form new roots (HI 1 32), for the desiderative and denominative 
taught here share many properties with the original roots.^^ The tapeworm 
develops, so to say, an enlarged head. The addition of further suffixes depends 
on the direction of growth: verbal or nominal. 

In a verbal form the element affixed next to the root is what we caU the tern- 
pus character, etc.; Katyayana and the later Paniniya-s call it vikarana 
'modification’ — ^but Panini has no general name for this class of suffixes.^o 
Before the personal endings of the first future an element -sya- is inserted after 
the root (e.g. dd-sya4i; III 1 33), before the endings of the aorist an element 
which is immediately replaced by the special characters of the different 
aorist classes {a-hdr-s-iti etc.; Ill 1 43—66). Before the personal endings of the 
present tense, imperfect, imperative, etc. (called collectively sdrvadhdtuka) the 
stem forming elements are added to (viz. classes 1, 4, 5, 6, 8) and inserted into 
(viz. class 7) the root: div-ya-ti, tud-a4i, etc. (Ill 1 68-81), but elided in the clas- 
ses 2 and 3 (11 4 72-75): ad-04i>at4i, etc.^i It is evident in aU these instances 
that the vikaraTm-s are dependent on the personal endings (which are endowed 
mth temporal or modal values) because this dependence is expressly laid down 
in HI 1 67 edrvadhdtuke . . . “Before a sdrvadhdtuka suffix . . 

P. Thieime, JAOS 76.2. 

3S The form bhuv-adayah is strange. As Patanjali (Mahabha^ya I 253, 2f.) points 
out we expect either (in continuous speech) hhv^ddayah or (in discontinuous speech) 
bhu^a^yah. But there are indications that for some grammarians bhuv-adayah 
womd be nomal sandhi (Abhayanandin on Jainendravyakarana 12 1; Bha?avrtti 
on Pa^ VI 1 77: Vyadi, Galava) ; then the expression could go back to Pardni’s 
predecessors ( Yroms-imn MImamsax, Samskrt Vyakaran ka itihas, I, p. 152f.). Of 

/v/ even in Sanskrit, e.g. habhuva. 
r formation of the present tense stem including its accentuation or the 
formation of primary nouns ffom them. 

^ personal endings only, the Mahabha§ya 

f whether vikamnu-a or personal endings denote the 

teminoloivl ’ action (active or passive meaning in Western 

the pereonal endings appears to be stronger— which 
leaves the vikarai^-s with no meaning at all. 

L Wth disappearing (2, 3) or alternative (4, 5) suffixes 

are inserted [B. Shekts, Grammatical method in Panini (New Haven, 1961), p. 2f.]. 



After this excursus on the verbal stems the joint presentation of verbal and 
nominal formations resumes. The three time- classes of reality (past, present 
and future) result in three sections for suffixes that have such a time connota- 
tion:^- bkute ‘past’ III 2 84, vartamdne ‘present’ III 2 123 and hhavisyati ‘future’ 
III 3 3. In the first section, e.g., the (nominal) zero suffix is added to the 
root ]/cz if it is joined with the noun agni: agni-ci^0=^t ‘somebody who has 
arranged the sacrificial fire’ (III 2 91),^ in the third section gamin is taught as 
‘somebody who intends to go’ (III 3 3), etc. The verbal suffixes stand side by 
side with nominal suffixes, but their tempus characterization is more differen- 
tiated. While the aorist suffix (LuA) is said to denote just past action, the 
imperfect ( La^) is more specifically noted for the past that precedes the present 
day, and the perfect (LiP) for the remote past. Even the present tense suffix 
(LaP) can be used for the past if it is accompanied by the particle sma. The 
category ‘future’ logically includes not only the future tenses (LfP and LuP) 
but also the modal forms: after all, the actions referred to in the imperative 
( LoP ) and optative ( Li^) all stiU lie in the future. 

The user has now at his disposal a wide array of suffixes which he can add to 
the root, both verbal and primary nominal (krt) suffixes- But the order to start 
the whole process of word and sentence formation, and the decision which of 
the suffixes to take, comes in the last section, the pivot of Panini’s grammar 
which returns to the basic concept of the hdraJca-s. “[If one wants] to denote 
the agent, [one can employ] a Z:r#-suffix” (Icartari Text m 4 67). But he can also 
use one of the X-forms (verbal personal endings, e. g. foca-ti) ; this is implied 
in III 4 69 [67 Tcartari] Lah harmani ca hhave cdkarmakehhyali “The verbal 
ending denotes [the agent] and the object and, after objectless^ roots, the 
existence [of the root meaning itself]. In traditional Western terminology, 

42 L. Benoit, JA 248.305-337. It is not quite clear what the three participles 
refer to. Patanjali (Mahabha^ya II 111, 2-5) considers first hale ‘for the past . , , 
time’ but rejects it because hale is not dittoed here. Then he suggests a locative 
dhdtau to be obtained from the ablative dhatoh in III 1 91 which is dittoed through 
the chapter. However as the roots themselves are not subject to the time categories 
he concludes that the actions denoted by the roots are meant. Another, and it 
seems better, possibility would be an anticipating reference to hartari, harmani and 
hhave in III 4 674-69. A logical difficulty can easily be met: gamin ‘somebody who 
intends to go’ will be an agent of going only in the future even if the person happens 
to be present; the alternative would be to accept a present agent of a future action 
which makes no sense at all. 

43 The same suffix is taught in HI 2 177-179 in a ‘present’ meaning. 

44 I.e. intransitive; the action and its result reside in the same person or thing. 

45 R. Roohee, in Recherches linguistiques en Belgique, ed. Y. Lebrun, 1966, 
p, 113—120 and La theorie des voix du verbe dans i’ecole panineenne (Bruxelles, 
1968), p. 23, and G. Cardona, Lingua 25.213-220, have drawn attention to the 
close relation of hhdva and hriyd\ but neither ‘etat’ nor ‘action’ are satisfactory 
translations for hhdva. hhdva in V 1 119 taaya hhdvas tva-ta^.au ^^~tva and -td [denote] 
the existence of this” cannot be left aside. ‘Existence [of the action]’ for hhdva has 
been proposed independently by A.Wezler, Bestimmung und Angabe der Funk- 
tion von Sekundar-Suffixen durch Panini, Wiesbaden 1975, p. 99—104. 


H. Scharfe • Grammatical Literature 

this would be the active voice, the personal and the impersonal passive. Of more 
limited importance are some of the following rules that rule in nominal suffixes 
to express the kdraka-s sampraddm, apdddm (hhl-ma is 'somebody of whom 
one is ajfraid’) and adhikaraom (HI 4 73”76). While we were made to believe 
that we could see the verbal 'tapeworm’ grow: root, root enlargement, vikarana 
and personal endings or ^r^-suffix, we see now that Panini held the worm by 
its tail aU along. That leads to the question what his grammar really is: analysis 
or construction (realization)? Does kartari hrt state that ‘a Ar^-suffix denotes 
the agent’4® or does it direct the user to 'use a Ar^suffix to denote the agent’ 
Panini’s procedure indicates that he favoured the second alternative; his sub- 
stitutions of roots, suffixes and sounds show that he did not start with the 
finished word or sentence but that these were the final products of the process. 
At the beginning stands the desire of a speaker to express himself: artha-nimiU 
taka eva iabdah "The sound utterance of course is brought about by the mean- 
ing” as Patanjah later says (Mahabhasya 1 114, 13 f. and III 253, 12-15). There 
has never been any doubt that the last three chapters, dealing with internal 
and external sandhi, are synthetical. B. Paddegon assumed a ^chotomy : the 
chapters 1 to 5 he labeled 'theory of the ultimate components of language, or 
the analytical part of grammar,’ the chapters 6 to 8 'the theory of phonological 
and morphological coalescense, or the synthetic part of grammar. Panini 
nowhere makes such a distinction— nor does Patanjah: the whole grammar is 
synthetic, so but its synthetic character is half hidden in the pecuhar presenta- 
tion which more or less follows the words as they emerge from the mouth of the 
speaker from root through vikarana and suffix (nominal hrt or verbal L) to the 
vibhakti-s. With the vibhakti-s verbs and nouns part them ways. For the L-ele- 
ment we substitute one of the 18 substitutes which correspond to the three 
numbers, three persons and the two genera verbi; in the case of certain tempora 
or modi, different forms will be substituted. The word components which have 
so far remained isolated are then joined in the latter part of the grammar 
through a great number of morphophonemic rules. The nouns too move from 
the abstract to the concrete: in IV 1 2 there are seven triplets corresponding to 
the seven eases (the vocative is regarded as a modified nominative) and three 

« 0. VON Boe^gk; “Bin hrt genanntes Suffix bezeichnet ... den Agens.” 

n. JrtENOu: La suffix hrt vaut quand it s’agit de Tagent ...” 

48 The combination of the morphological elements into words and the morpho- 
pnoneinic mteractions between words in a sentence. ^ 

/’f granamar (Amsterdam, 1936), p. 61 and 53 ; 

cf. ^ H.E. Buiskool, The Tripadi (Leiden, 1939), p. 15-17. 

Noqten does not free himself 
an analytical/synthetical dichotomy. The sutra anyebh^o 
driyate [ suffix ... is also found after other [roots]” (HI 2 178 ; 3 130)1 and some 



numbers, 21 suffixes in all that are added to the noun in accordance with the 
rules n 3 1-73 {MraZ:a-case-relations). Again the coalescence into words is 
treated later, viz. in the last three books (together with the final development 
of the verbal forms as many morphophonemic rules apply equally to nominal 
and verbal forms). 

Let us take as an example a simple fact and its expression in the Sanskrit 
language: Devadatta sings a praise (hjman). Besides the proper name Devadatta, 
there are the roots ‘sing’ and ]/S^ ‘praise.’ If the speaker wants to take 
‘singing’ as the basic or connecting notion and if he chooses an agent-directed 
construction, then the form gdy-a-ti denotes not only ‘singing’ but also the 
(present!) agent in the singular ‘he/she/it sings’ ; the nominative Devadattah only 
fills the abstract notion ‘agent’ with a concrete personality; the object of sing- 
ing, viz. the praise, appears in the accusative:^^ Devadattah stotram gdyati “De- 
vadatta sings a hymn.” In the goal-directed construction, gl-ya4e denotes 
(besides ‘singing’) the (present) object ‘he/she/it is being sung’; the no m i n ative 
stotram only details the specific item. The agent of singing appears in the instru- 
mental: Devadattena stotram glyate “A hymn is being sung by Devadatta.” But 
the speaker may as well decide to express ‘praise’ by the verb and ends up 
with a sentence Devadatto gitena stauti “Devadatta praises with a song;” or he 
may choose a noun construction like Devadattasya stotra-gdyanam . . . “Deva- 
datta’s praise singing ...” The word stotram in the sentence Devadattah stotram 
gdyati contains two hdraha relations: the ^r^-suffix 4ra- denotes the Jcarana 
‘instrument’ of the root notion, i.e. ‘instrument of praising’ (by rule III 2 182), 
the accusative ending the harman ‘object’ of the action of ‘singing’ expressed 
by the verb.^s 

Nominal composition, though taught much earlier in the grammar (II 1 1—2 
38), really enters only now into the grammatical process. The rule that releases 
this mechanism is II 1 1 samarthah padavidhih “[Now] the finite-word rule, with 
unified object:” when words are so close that they together refer to a single thing 
or notion, composition replaces case suffixes.^^ An example is rule II 1 37 
pancami bhayena “[A word ending in] an ablative suffix is compounded with 
hhaya'' e.g., vrhdd hhayam ‘fear of a >vrJca4)hayam ‘wolf-fear.’ 

51 Different from oxir system, number is not the primary classifier: the three 
nominatives (sing., dual, plural) come first, then the three accusatives; similarly 
among verbal endings: the three third persons, the three second persons, etc. 

52 stotra ‘praise, hymn’ is derived from the root '^stu ‘praise’ with the suffix -tra 
expressing the A;amA;a-function ‘instrument’ (III 2 182): ‘instrument of praising ; 
this complex notion becomes the object of ^gai ‘sing’ and receives therefore the 
accusative case ending. 

53 In Patanjali’s examples (Mahabha^ya I 442, 7f.) prdsdda dste; say ana aste He 
sits on the terrace ; he sits on the bed” the suffix -ana denotes adhiJcarana ‘location’— 
and so does the locative suffix. Here -ana denotes the location of ysad ‘sit’ and ysi 
‘lie,’ respectively; the locative, the location of yds ‘sit.’ 

54 Patanjali’s discussion of this important sutra has been translated and explained 
by S.D. JosHi, Patanjali’s Mahabhasya, Samarthahnika (Poona, 1968). 


H. Scharfe • Grammatical Literature 

Most of the fourth chapter and all of the fifth are devoted to the secondary 
noun suffixes (taddhita)^^ which optionally (IV 1 82, viz., instead of an otherwise 
indicated compound) replace longer analytic expressions when the condition 
of a unified concept is Examples are the patronymics (IV 1 92-178): 
Dak^asydfatyam ‘Daksa’s descendent" >Daksa+i^>DdJm (IV 1 95), etc. The 
number of these secondary suffixes is great, and the variety of their meanings 
shows the sophistication of the language. The Jcdraha-velsitioiis play hardly any 
role, and even that only indirectly: V 4 45 [42 anyatarasydm 44 pancamyds 
apaddne ca . . . may be substituted for the ablative suffix if the 
suffix denoted apaddm . . The rules about taddhita-mf&xe^ and composi- 
tion presuppose a completed ‘basic language' with verbal and nominal endings 
and primary noun formation; for ‘unified concepts'^? they offer abbreviated 
expressions. DdJcp points to one certain person only, while Dahsasydpatyam 
refers to both Daksa and the person descended from him. 

From the sixth chapter to the end of the grammar Panini lays down the 
mechanics that create words and sentences^® out of the morphemes taught so 
far:^® where similar vowels meet they merge in their corresponding long variety 
(e.g.i+i>i), in other cases we have diphthongs (e.g. a + i>e), etc.; consonants 
are also variously affected. A large segment of the sixth chapter (VI 1 158-2 199) 
deals with the three pitch accents as they result from suffixation and composition. 

55 These suffixes form nouns from other nouns. Some of them are first taught 
with heterophonic elements which are later replaced with the actual morphemes: 
YU and VU stand for actual -anct and •aha, respectively (VII 1 1), PH, DH, etc. 
stand for -dyan, -ey, etc. (VII 1 2) ; cf. H, Sohabfb, Panini’s metalanguage, p. 24f. The 

• ^ y i i X / vuu-urnam JL ne suinx -vat} is usea 

to denote fittmg for x.’ ” But in Panini’s grammar the corresponding rule reads slightly 
dmemntly V 1 5 tasmai hitam “good for it” and does not occupy a prominent place. 
\Aj WEa^, Be^immung und Angabe der Funktion von Sekundar-Suffixen durch 
Pa^, Wiesbaden 1975, p.l39 proposes instead for taddhita a translation ‘daran 
gefugp (= ‘joined to x’).] 

56 samarthdndm . , . vd of IV 1 82 refers back to II 1 1 samarthah padavidhih of the 
^apter on compounds. On the traditional interpretation of Pan TV I 82 cf. A. 
Wezleb, Bestimmung, passim. 

_ S' To mention jiMt one other group of derivatives, the suffix THaP ( =iha) is 

^ pkvrfika ‘born in 

tne rainy season (IV 3 25f.) .c' . v 

‘in ^ ^ 2 108) which means 

^ (definition 1 4 109) . This is not necessarily 

117 but often a sentence phrase fvargaj : H. Sckame, ZDMG 

-ft’’ und Sohlussvokalverzeichnungen in A6oka- 

(JBOES n ®^®’^'^®®'^®snin’s commentary on ArthaSastra II 10 

^ Josm IL 2fi andlTddyota on I 3 1 (beginning) and 

stm +i« andpravr? + 



For the final section of his grammar Pamni creates a one-directional string of 
rules, i. e., each rule here applies to forms as they appear up to this point but is 
‘non-existent’ ( asiddha)^^ when it comes to the application of any earlier rule. 
This is the so-called tripadi ‘The Three Sections’ (VTTT 2 1-4 The tripddl 
deals mainly with special accent rules and with consonant sandhi. The linear 
nature of these rules prevents them from becoming the cause for the apphcation 
of an earlier rule: when ^rdjan looses its final /n/ before the suffix -hhis by VIII 
(> rdja-hJiis ) , this rules does not cause the application of VII 3 102 (lengthen- 
ing of final I a. I before suffix) or VII 1 9 (stems ending in /a/ take the suffix -ais 
instead of -bkis in the instrumental plural) — ^because the rule VIII 2 7 is ‘non- 
existent’ with regard to these preceding rules. The very last rule (VIII 4 68 a a) 
corrects an assumption made throughout the grammar (cf. p. 92). 

Tradition attaches several appendices to the Astadhyajn, yet not all of them 
appear authentic when scrutinized closely. The first and foremost appendix is 
the root list (dhdtu-pdtha) which contains aU roots of the language arranged in 
ten classes — dependent on how their verbs form their present tense stem;®^ 
roots whose verbs can follow one or the other formation are listed under both 
classes (e. g. ]/mr/ in classes 2 and 10). Panini refers in his grammar to these lists 
with references Hke div-ddihhyah (in III 1 69) “After the roots ydiv, etc.” In 
six instances he refers to a certain number of roots from such a class, e. g. VII 
3 98 rvdas ca pancabhydh “After the five roots ]/md, etc.” (i. e. ^rud and the four 
roots that follow it in the root list). Panini was not satisfied merely to list the 
roots according to their present tense classes ; through various attached deter- 
minatives and technical accents he marks them for active or middle inflection 
of their verbs, insertion of /i/ between the root and certain suffixes, loss of a 
final nasal, etc. These additional markings are conclusive proof that the dhatu- 
pdtha is authentic, for Panini teaches these markings in his sutra-s (e.g. du in 
13 5 and III 3 88) while their applications occur only in the root list. Occasion- 
ally Panini wants to grasp in a sutra several roots that even with aU these 
devices cannot be called upon summarily; in these cases he has no choice but 
to list the roots in his sutra one by one (e.g. Ill 2 182). Though the dhdtu-pdtha 
as a whole is proven old, this cannot be said about many single roots: there are 
many apparent doublets that may have entered into the text at a later date 

Pa nini has invoked the asiddha-pTmaiple also in VI 1 86 and VI 4 22, though 
on a more limited scale. 

61 This part of the A§tadhyayi has been studied in an exemplary fashion by 
H.B. Buiskool, Purvatrasiddham (Amsterdam, 1934) ; an abridged English recast 
of this study is The Tripadi (Leiden, 1939). 

62 This arrangement is easily justified by the statistical preponderance of present 
stem forms. 

63 yuch puck much I 229 and jhas ws I 720 are graphic variants due to the simi- 
larity of the first letters in some script or other. Extensive commentaries on the 
Dhatupatha are the Dhatupradipa of Maitreyarak§ita from Bengal and the K^ira- 
tarahgini of KIsirasvainm from Kashmir (both 12th c. A.L).) and the Madhaviya- 
dhatuvrtti of Sayana (14th c. A.D.). 


H. Scharfe • Grammatical Literature 

the Pamniya dhdtu-pdtha is less well preserved than the root lists of some later 
grammatical schools. Patahjali has on several occasions proposed adding a root 
or changing a technical accent®^ and even where he finally rejected the amend- 
ment it has found its way into the traditional text. A major addition to the 
dhatu-pdiha is the root meanings which we now find after each root in the 
form of a noun in the locative case, e. g. II 1 ad® hhaksane ^]/ad in the meaning 
of eating.’ When Patanjali once (Mahabhasya 1 254, 12) recites a segment of the 
root lists, he recites the roots only: hhv-edh- (cf. the reading of the traditional 
list bhusaUdydm; edh^ vrddhau; . . . with the meanings added after each root). 
Patanjali may however already have known a list of root meanings, because he 
contrasts in Mahabhasya I 256, 11-15 certain standard {drsta 'noted’) root 
meanings with other occurring meanings.^s Tradition attributes the addition 
of meanings to a certain Bhimasena^s who probably only codified traditions 
current at his time. The modernity of the root meanings is still recognized by 
Bhattojidiksita (17th cent.).®^ 

A second appendix consists of the over 200 word-lists (gana), often collec- 
tively called gana~pdtha. These Hsts do not form a corpus but occur in the 
commentary on the sutra-s, e. g. in the Kasika-vrtti, wherever a sutra with an 
expression like svar~ddini ('smr, etc.’) called for such elaboration. It is generally 
conceded that the lists were expanded over the centuries, not only through 
copying mistakes but also through deliberate additions; the occurrence of a 
word in the gana-patha is therefore no proof that the word or its object were 
known to Panim.®® The insertion of new words into the list was facilitated by 
the doctrine (first indicated by Patanjali, Mahabhasya I 400, 13) that besides 
‘complete’ lists (sampuTTm-gaTm) there are lists that comprise all words of the 
same type (dhriugaim) even if these are not expressly mentioned in it. As the 
need to insert new words continued, more and more gaim-^ were gradually 
claimed to be such akHi^gams: gana 219 vrihy-ddayah on Panini V 2 116 was 
not regarded as an dhrii-gaiui in the Kasika, but later was so regarded in Va- 
mana’s Kavyalamkara-sutra V 2 57. The distinction is not based on any 
mdication of Panini’s who uses in his sutra-s references like amsv-adayah, urah^ 
prabhrtaya^, gavdsva^prahhrtlni gotrddini and ardharcdh with no apparent 
diffemnce in meaning. In a few instances [ayasrmyddlni 1 4 20, indrajanandda^ 
yah Iv 3 88, tujddayah VI 1 7) the vrtti has no list at aU. Besides such words the 
gam^s contain a number of sutra-like sentences which the later Paninlya-s 

fil^sf ^ Patafijah, Yale thesis, 1963 (University Micro- 

Z Sanskrit Dhatupathas (Poona, 1961), p. 93f. 

Earlier th^ A.D. 600 (P.K. Gode, mA 2. 108-110) ; already Gandragomin pre- 
supposes the addition of the root meanings ^ ^ gominpre 

3210 ( =PSn m I 47). The frequently givea meaaiag gaiau 
meaaiag; cf. G.BtefwzS g.S ^ 

S. Sengupta, JAS (Calc.) 3.89-186 and S.M. Ayachit, IL 22. 1-63. 



call ga 7 Ui-sutra-B ; many of them are identical with sutra-s of Panini or varttika-s 
of Katyayana while the origin of others is unclear. 

What is the evidence for Panini’s authorship of the The question 

must he asked because a sutra like 1 1 27 sarvddini sarvandmdni ‘'Pronouns are 
sarva, etc.’’ does not necessarily indicate the existence of a list 'sarva, etc.’ It 
is hkely that Panini explained the rule to his students and named other words 
that fell under his term sarvaTidinan, hut he need not have given them a fixed 
hst, on a par with the sutra-s themselves or with the dhdtu-pdtha. Six times he 
refers to the dhdtu-pdtha with a numbered reference (e.g. VII 2 75 Jciras ca 
pahcabhyah “And after the five roots ]/^r, etc.”),'^® twice to the sutra-s (VI 2 135 
sat ca hdndddlni “And the six words Icai/^da, etc.” refers to words mentioned in 
the rules VI 2 126-129; VII 1 16 pmrvddihhyo navabhyo vd “Optionally after the 
nine words purva, etc.” refers to the nine words mentioned in 1 1 34-36). But 
only once Panini seems to refer to the gayu-s in this way: VII 1 25 datard- 
dihhyah pancahhyah “The ending -at is substituted after the five words ending 
in -atara,^'^ etc.” which includes words ending in -atamaP^ as well as itara, anya 
and anyatara — altogether 9 words. Is this a reference to a section of the gaim 
sarvddini (241, 5-9) or did Panini merely want to indicate that five analogous 
formations or words form the neuter nom./acc. sing, in -at and left it to the stu- 
dent to identify these five? Twice, it was mentioned, Panini refers with num- 
bered references to words taught in the sutra-s: VI 2 135 sat ca hdr^ddlni and 
Vn 1 16 puTvddihhyo navabhyo vd; if there were fixed gaTia-s there is a danger 
that the first would be taken to refer to kdnda in the gana bilvddayah (thus 
denotiug the six words kdT^a, mudga, masura, godhuma, iksu and venu), the 
second to purva in the gaTm ardharcdh (and thus denoting the nine words purva, 
camasa, ksira, karsa, dkdsa, astapada, mangala, nidhana and niryasa). The diffi- 
culty disappears if there were no fixed gaTia-^ in Panini’s time. 

If the fixed gan/i-s should go back to Panini himself, it remains unclear why 
we have so many sutra-s with long lists of words (e.g. II 1 65; III 1 21) instead 
of a short reference to an attached ganiai Vedic nipdtana rules (which teach the 
desired forms directly) almost never use the gra^-technique to reduce their 
buMness (e.g. Ill 1 123).’^ When a quotation of roots cannot be given with a 

An extreme position was taken by B. Birwe, Ber Ganapatha zu den Adhyay^ 
rV und V der Grammatik Paninis (Wiesbaden, 1961), who on p. 18, fn. 1, attributes 
most ga 7 Wb- 8 utra-Q to Panini and on p. 26 f. some to his predecessors because of 
their ‘pre-Paninian terminology:’ dkhydta ‘verb’ in gana mayuta-vyarnsahadayali, sa 
‘compound’ (short for samdsa) in gaTmutsadayaho^ndsarndhy-aksara^^pYithong' in 
gana svar-adlni. This argument however is inconclusive because dkhydta and satndhy- 
aksara were used by Katyayana and other Panimya-s, sa in the late Jainendra- 

The other instances are III 2 141 ; VI 1 6 and 4 125; VII 3 74 and 98. 

I. e. katara, yatara and tatara (V 3 92: ^atara^). 

72 I.e. katama, yatama and tatama (V 3 93: ^atuma^). 

73 L. Renotj, Etudes vediques et pamneennes I (Paris, 1955), p. 109; sole excep- 
tion is VII 1 49 snatvy-adayas ca. 


H. Scharfe • Qrammatical Literature 

reference to the dhdtu-'patha^'^ because the roots are not consecutive in it, 
Panitii lists the roots in the siitra (e. g. I 2 7 rnrda-mrda-gicdha-kusa-Uiia-vada- 
ms,ah Hvd) without recourse to a gana. In Panini’s treatment of the particles 
( nipdta) the only element of vagueness is in the sutra 1 4 57 [56 ni^tdh] cddayo 
^sattve ''ca, etc. [are nipdta-s] except when they denote substances*': the defini- 
tion of "ca, etc.* depends on the gam attached to it. It is worth noting that the 
Brhaddevata 11 93 states that iyanta iti sarnhhydnam nifdtdndm na vidyate 
“There does not exist an enumeration of the particles stating explicitly: ‘there 
are so many"*; almost identical is the statement of Rgveda-prati^akhya XII 9 
(26) niyanta ity asti sarnMiyd. Shall we assume that the gams were non-existent 
when these two texts were composed? This would force us to assume for them 
a higher date than for Katyayana who was quite familiar with the ganas. Or did 
the author(s) of the two texts ignore this part of Panini’s work deliberately? 
The likely answer is that the gam cddayah was regarded by them as an dkrti- 

The gams lack a feature typical for sutra-^ and dhatu-patha: Papini’s knack 
of putting every formulation to double or triple use through- a resourceful 
employment of sequence, accents and determinatives. The gapus may in 
character if not in the exact formulation represent the explanations given to 
students. If that is so, then even the insertion of sutra-s into a gam (e.g. I 1 
34-36 in the gam sarvddini) is justified: as a mere duplication it cannot claim to 
teach anything new but it assists in the interpretation of the sutra I 1 27 
sarvddini sarvandmdnL It is remarkable that Jinendrabuddhi (11th cent.) in his 
Xyasa on the Ka^ika on IV 1 106, V 3 2 and VII 4 3 claims that the gams are 
not Panman and that the author of the gavLas is not the author of the sutra-s 
(anyo hi gana-kdrah, anyas ca sutradcdrdh) . 

A third aUeged appendix is the u’^-adi-sutra-s’’^ in which nouns are derived 
from roots in a more irreplar fashion. Panini twice (III 3 1 and 4 75) refers to 
it’, etc., but it is not clear whether he referred to a specific list and 
it he did, to which or whose. The uv-oiit-list which is now attached to his gram- 
mar (e.g. in the appendix to the Siddhanta-kaumudi) is sometimes ascribed to 
bakafayana.” Because Sakatayana, Pa^iini’s forerunner, held the view that 
eve^ noim can be denved from a root he must almost of necessity have had a 
Himilarhst of irregular suffixes; the Paniniya-s, on the other hand, believe that 

five ^ “Also after the 

Note also that Nirukta 14-11 gives a list of 22 nils Z Zi^s no complete- 

The UnS^to^s of XJj jvaladatta by Tn. Aufiiecht (London, 1859). 

patha and pariiistas (Poona 19^)^ 724-7^^7 
ed. K.K. R aja (M adL. 1956) L Rbhou JA 
” E.g. S. Paxhak and S. cLbao, Cd Mai! p ^ 



the words with u^-ddi suffixes are really not built up {uV'-ddayo ^vyidpanndni 
pratipadikani, parihhdsd 22 in Nagojibhatta’s Paribhasendusekhara). The ques- 
tion arises why Panini referred to u^-adi suffixes at all if he disapproved of the 
procedure. It is perplexing, to say the least, that Panini often refers to suffixes 
which he has never introduced. Some of these are mentioned without determina- 
tives, e.g. as-anta ‘[stems] ending in -os' (VI 4 14), 4u- (VII 2 9) or is-us.oh ‘of 
[words ending in] -is or -us" (VIII 3 44) ; one could possibly argue that /as/, /is/, 
/us/ and /tu/ were not conceived as suffixes. TMs is however not possible for 
VI 4 97 is-man-tra^-^ ca “Also before [the suffixes] -is, -man, -tra and 
ZERO” where the suffixes -is (chadis), -man (chadman), -tra (chattra) and 
ZERO (upacchad) are mentioned side by side: tra^ has a determinative it 
cannot be a reduced form of previously taught Hra^ (III 2 181-186) because the 
sphere of Hra"^ is restricted to certain roots excluding ^chod. As Panini has never 
taught the use of these suffixes one must conclude that they were w^-adi suf- 
fixes; and uV' are quoted with their determinatives, others (e.g. -is, -us and 
-tu) without them. That corresponds to Panmi’s frequent practice to refer to 
his own suffixes in a short form. While a suffix is^ is found in our (spurious) 
U^i-adi-sutra-s to account for chadis (265), chattra is not accounted for. If 
Panini knew some u^-ddi lists, there is no indication that he knew the Un-adi- 
sutra-s we have. In fact Un-adi-sutra 73 cay ah kih gives every impression of 
being modelled after Panini VI 1 21 cdyah ki,’^^ Un-adi-sutra 215 *VI^ vaci- dlrgho ^ samprasdranam ca after Katyayana’s vart- 
tika 2 on III 2 178 vaci-pracchy-dyatastu-kapipru-ju-srlimm dlrghas ca, Patan- 
jali’s suggestion (Mahabhasya II 135, 14) hhiyah ^ruka^ api vaktavyah “The 
suffix -ruka- after also must be taught (to account for hhlruka ‘fearful’)” 
was apparently the source for Un-adi-sutra 189 hhiyah ^ruka^J^ I conclude that 
Panini knew some uV'-ddi list but that the ones we have are later than Patanjah. 

The Linganu^asana ‘Instruction about gender’^o sometimes attributed to 
Panini (e.g. Appendix to the Siddhanta-kaumudi) is probably much later than 
Patanjali. It is not mentioned in the Mahabhasya and Patanjali’s thesis lingam 
aHsyani lokdirayatvdl lihgasya “The gender need not be taught because it is 
based on usage” (Mahabhasya 1 390, 18f. and often) speaks against the existence 
of an authoritative work on gender in his time. The composition of gender man- 
uals was probably inspired by the lexica which often have a special section on 

The Phit-sutra-s^i of ^antanava are also probably later than Patanjah. They 
teach in a mechanical way the Vedic accents of noun stems. 

Cf. also Panini VI 2 139 and Unadisutra 666 which in spite of a striking simi- 
larity are materially incompatible. 

The Moka quoted Mahabhasya I 36. 8 corresponds to Unadisutra 350. 

80 0. Fbaitke, Die indischen Genuslehren (Kiel, 1890). 

81 F. Kxblhoknt, Phitsutrani, Qantanava’s Phitsutra (Leipzig, 1866; reprint Nen- 
deln, Liechtenstein , 1966); ed. G.V. Devasthali (Poona, 1967). 


H. Scharfe • Graromatical Literature 

Panini seven times refers to the language' (bhasa\ e.g. Ill 2 108), in contrast 
to the language of the older Vedic literature (chandas).^^ Several of his rules 
refer clearly to spoken language: in VIII 2 83 f. we are told that the last vowel 
in a sentence has the high pitch accent and is extra long when one responds to a 
greeting — except when the other man is a iudra (low caste man) ; the same 
applies if one shouts from a distance. But Panini is concerned only with the 
spoken idiom of the educated classes; the language of the lower classes was, in 
his time, probably much closer to the Prakrit dialects as the occurrence of 
popular forms in texts as old as the Pgveda (e. g. ahJikJialikrtya) and the promi- 
nence of Middle Indo-Aryan dialects at the times of Buddha and Ai^oka 
Priyadar^ indicate. 

Panini’s verbal paradigm had lost many of the competing options we find in 
the older Vedic texts and verb and prefix are now joined together; in noun 
inflection, the ancient dLstinction of devl and vrU tjpes is a thing of the past. 
On the other hand, the distinctions of the past tenses (perfect for the distant 
past, aorist for the most recent past, imperfect for the past before the present 
day) are still retained, the verbal duals in ~dte and -dtlie are not yet pragrhya and 
the subjunctive is still taught. The periphrastic perfect (unknown to the oldest 
Vedic texts) is formed only with ]/A;r ; coraydin cakdra in accordance with later 
Vedic usage. Panini's language is virtually the language of the brahmana-s 
and sutra-s of the later Vedic 

Considering that Pamni hailed from Salatura in the northwest Punjab it is 
not surprising that his language is especially close to that of the Kathaka- 
sam^ta which represents the northern Yajurveda tradition as opposed to those 
traditions of the East (Vajasaneyi-samhita), South (Taittiriya-samhita) and the 
West (Maitrayani-samhita):^^ amdvasyd besides amdvdsyd (III 1 122), apdm- 
naptriya (IV 2 28) and hatipayatha (V 2 51).85 The syntax of cases fits closely 
with the Vedic prose of the brahmana-s and sutra-s.®® The language of the North 
was regarded the best in Kau^itaki Brahmana VII 6 tasmdd udlcydm disi 
prajTwiaUird vdg ucyata udanca u eva yanti vdcam Mksitum yo vd tata dgacchati 
tasya vd susrusante “In the northern region, speech is spoken particularly 
distinct. So, people go to the North to learn speech. Or, if someone comes from 
there, they like to hear (learn) from him." Panini's grammar must have acquired 

I references to hhm. 

^ concluded that the bhd0-rules were unimportant. But 

Ve^fn basic language description after an excursus on 

^ Etudes vediques et pamn^ennes I, p. 114f.). 

vprh perf. m and in ••ana may still be used instead of a finite 

verb form, not o^y m the Veda (HI 2 106^107) but sometimes even in the current 
language (bhasa; lll 2 108-109): P. Thieme, KZ 78. 95. current 

(Leipzig, 1881), I, p. XIX-XXVIII; 
iogicrScat^!l54 (Gottingen, 1923), p. 301f.=Philo- 



its position of authority at a time when the language of the North was yet felt 
to be exemplary. Later the highest authority on language and customs rests 
with the people of Aryavarta (central North India): Mahabhasya III 174 
7 - 10.87 

Panini’s credibility was severely challenged by W.D. Whitneyss who pointed 
out the great number of unattested roots in the Dhatu-patha (he found only 
little more than one third of its roots attested in other than grammatical texts) 
and other peculiar forms taught in the Astadhyayi but not found in literature. 
We must consider however that we have only fragments of the Vedic hterature 
and that Panini, along with such texts, rehed on the spoken language that was 
fandliar to him. It is only natural that he knew and taught words and forms 
that have not come down to us. This became apparent when subsequently some 
such forms came to light in L. von Schroeder’s editions of the Kathaka and the 
Maitrayani Samhita. Concerning the unattested roots, one must further keep 
in mind that the tradition of the Dhatupatha is much less rehable than that of 
the Sutrapatha ; it is easier to insert items into a list and moreover the Maha- 
bhasya affords us hardly any checks on the tradition of the root lists. The exist- 
ence, side by side, of the roots hakh (1 124 and 821) with the variants kakkh and 
kJiakh (1 124) and ghaggh (1 170) with the variants gaggh and ghagh all meaning 
‘laugh’ points to onomatopoetic formations in colloquial usage; similarly catc, 
jarc, jarts, jharjh ‘rebuke, chide’ which remind of tarj and hharts may he lo- 
cal variants. Other roots have only nominal and no verbal derivatives, 
which still makes them ehgible for the root lists; the notion of ‘verbal’ 
roots is not Panioian. It may be true that these roots have been set up to 
account for the formation of a few nouns: they are ‘artificial,’ fictitious, e.g. 
]/^ (I 985) accounting for gJirta, ghrnu and gharma,^^ But so are all of Panini’s 
roots ; they are abstracts, attempts to pronounce mental images, and he never 
entertained the notion that these at some time past formed separate words 
with suffixes following. Panini thought strictly synchronically, not historically. 

Harsh criticism was also directed at the Vedic rules in Panini’s grammar. It 
was felt that here we have, in contrast to the spoken language, the control 
material to test his competence. On the one hand Panini notices the finest 
distinctions: he has observed that the abhinihita-saTndhi, i.e. the elision of an 
initial /a/ after word final /e/ or /o/ is not observed inside a verse line unless /a/ 
is followed by /v/ or /y/ (VI 1 115) — and the next rules give seven exceptions to 

87 Cf. Baudhayana Dharmasutra 1 1, 2, 10 and Vasi^tha Dharmasastra I 8-9. 

88 W.D. Whitney, AJPh 5.279-297; 14.171-197; GSAI 7.243-254. 

89 Mahabhasya III 275. 15f., cf. G.B. PAnsunE, The Sanskrit Dhatupathas, p. 
198 and 208 f. On what basis these roots were assigned to verb classes and why they 
received determinatives is not clear. But this problem is not confined to roots like 
}/ghri drs that has no present tense stem is Hsted with the first class (I 1037) as are 
also some other defective roots without present tense stems that are taught as root 
substitutes (vadh for han, ah for bru ) . In some cases root substitutes in present tense 
stems {yach for da, posy for drs, dhau for sr, etc.) are not listed in the Dhatupatha. 


H. Scharfe • Grammatical Literature 

the exception. All these statements check out perfectly with the Vedic texts. 
On the other hand, the presentation of the Vedic peculiarities is very spotty. 
Major features are not treated at all and the selection of archaisms seems to he 
without principle.Qo It is clear from this that the Vedic rules have only the 
character of a supplement; moreover, it was not necessary to repeat the 
treatment of features that the Vedic dialects had in common with the later 
language. A method of description somewhat strange to the spirit of Panini’s 
grammar is more frequently employed for Vedic forms, viz. mentioning of the 
finished form without any grammatical hmld-up (so-called ni'pdtana-sutra-^\ 
nipdiana lit. ‘letting it drop in’) sometimes a whole Vedic passage is quoted 
(e.g. Vni43;2 69). 

Several times Panini refers to the Eastern or Northern usage the genitives 
prdcdm and vdicdm need not refer to Eastern or Northern grammarians as 
the Katika on IV 1 17 suggests (^prdcdm dodryandm matena). Being himself 
from the North, Panini gives more intimate details concerning Northern usage: 
IV 2 74 informs us that the names of some wells north of the river Vipa^ (modem 
Bias) vary in their accent from those south of the river (a well dug by a man 
named Dalta would be called datta north but ddttd south of that river). Other 
Northern expressions were probably rare archaisms like rmtara-pitarau ‘mother 
and father’ (VI 3 32) against conamon mdtd-pitarau or pitarau (I 2 70). The 
‘Eastern’ rules refer to Eastern cities, their inhabitants and to features of their 
speech. Of these the word ekatama ‘one of many’ (V 3 94) is actually first 
attested in an Eastern text, the iSatapatha-brahmana 1 6, 3, 23. 

Panim’s object was the colloquial as well as literary language of the educated 
without any reference to certain Vedic branches of tradition. Once his work 
was classified as a ‘limb of Veda’ (veddnga) it became necessary to say this, 
because ordinarily Vedic traditions were the property of certain schools alone. 
Patanjali defends the formulation of optional rules by saying: surva-veda- 
pdrisada/tTi hidcLin idsttam* tatra naihdh pantlidlh iahyd dsthdtuM “This instruc- 
tional work belongs to all Vedic schools; therefore it is not possible to resort to 
one single way” (Mahabhasya I 400, lOf.; HI 146, 15f.). PanM’s rules were 
accepted, we do not know since what time, as authority on Sanskrit grammar 
wherever Sanskrit was used; later remakes were concerned almost exclusively 
with easier or more logical presentation and only rarely with new observations 
of usage. That does not mean that everything Panini taught was put into 
practice. The rule VUI 4 56 vdvasdne states that in final position with no other 
word following immediately (i. e. at the end of a sentence or sentence segment) 

• noted archaisms occur in Veda sections that played a larger role in 

pnestly practice? 

^^diques et panin^ennes I, p. 103-114. The technique 
prob^ly derives from lists of rare Vedic words (cf. the Nighantu). It explains the 
aJmo^ c^plete absence of Vedic gana-s (sole exception VII 1 49 snatvy-adayas ca). 

Cf. Mahabha§ya I 105, ^13. 



voiced stops are only optionally devoiced: suhrd besides sukrt.^^ Later gram- 
raarians paid lip service to this rule (e.g. the Ka^a on this rule; Candragomin 
VI 4 149 ; Jainendra V 4 131) but the voiced stop is never found in this position 
in any texts as far as oux manuscripts go.®^ 

Pa riini and with him the grammarians that contributed to the science of 
grammar before him, owe their greatness to a combination of fundamental 
discoveries: 1) the insight that the proper object of gram m ar is the spoken 
language, not its written presentation; 2) the theory of substitution; 3) the 
analysis in root and suffix; 4) the recognition of ablaut correspondences; 5) the 
formal description of language as against a logical’ characterization; and 6) the 
concise formulation through the use of a metalanguage. It is often said that the 
transparent nature of Sanskrit made the analysis possible. But we can argue 
as well that it was first Panini’s (and his predecessors’) analysis which made the 
structure so transparent: was the relationship of dohmi and adhuksat, or majjati 
and madgu reaUy that obvious ? 

While the Pratisakhya-s and our own popular grammar have it that e.g. the 
sound /b/ becomes /p/ under certain conditions, it has always been the view 
of Indian grammarians that the sound /p/ is substituted for /b/; /b/ does not 
change, it is our mind that switches from one phoneme to the other. Before 
certain suffixes, the root hhu is substituted for the root as (II 4 52) ; here it 
would be difficult anyway to speak of 'change.’ Substitutions apply to single 
sounds, sotmd groups, suffixes and roots. Substitutions make it possible to 
derive the parallel infiections of different noun classes and the pronouns from 
a single set of case suffixes but no attempt is made to farther reduce the case 
suffixes of singular, dual and plural to a single set with separate number mar- 
kers because no Sanskrit paradigm shows such a reduction. It is necessary that 
the substitutions are valid throughout unless limited by special conditions. 
Strict application of the theory led to remarkable results of internal reconstruc- 
tion. The sentence variants asms, asvah, asvas, asvo, asm, etc. can be explained 
from a basic form asvas; this happens to be historically the oldest form (I.E. 
*e% 05 ). Even more striking is the assumption of a root mm] as the common 
basis of the verb majjati 'submerges’ and the noun madgu 'diver-bird’ which 
can be derived from it by independently established substitution rules. The 
comparison with Latin mergo 'submerge’ and m&rgus 'diver-bird’ leads us to a 
reconstructed Indo-European root mesg 'dive:’ Pa n ini’s internal reconstruction 
could not have been closer. 

93 Rgveda-prati4akhyal 3 (15/16) teUsusthat Gargya favoured the voiced, Sakata- 
yana the unvoiced stop ; Caturadhyayika 1 8 recognizes only the unvoiced as correct. 

94 W.D. Whitney, S anslmt grammar, § 141b ; J.WACKEiiHAGELjAltindischeGram- 
matikl, p. 302f. §260a. 

The case suffixes taught in IV 1 2 are essentially those that appear with the 
noun stems ending in consonants ; but the nom. sing, suffix 5^ is^ only reached by 
internal reconstruction as it never appears with the above-mentioned stems. The 
internal reconstruction is confirmed by comparative and historical reconstruction. 
Sanskrit vak, Latin vox (i. e. vok-s). 


H. Scharfe • Grammatical Literature 

At a later time the substitution theory came into conflict with the Mimamsa 
dogma of the permanence of words/speech sounds (iahda-nityatm): how can 
the sabda-s be permanent if their constituent elements are subject to substitu- 
tions? Katyayana who raised the question in his varttika 12 on 1 1 56 answered 
in varttika 14 that it is only our notion of sounds that undergoes substitution. 
In deference to the Mimamsa finally a compromise was reached that left both 
concepts outwardly intact but warped: all of Panini’s substitutions are sub- 
stitutions of complete words ; in a series of substitutions asva + su > asve + sii > 
asvesu we have not just a substitution of a>e and s>s (nor do /a/ and /s/ 
'change" to /e/ and /s/) but first asvesu is substituted for ahasu and then asvesu 
for ahesu.^^ 

The analysis in root and suffix went beyond a mere separation of 'word" and 
'ending" or a mechanical division of root and the remainder of the word. To a 
meaning-carrying nucleus called 'root" a well ordered string of suffixes is added. 
Some of these suffixes called krt can only be attached to roots directly, others 
called taddhita can be added only to such previously established composites, 
sometimes a few in a row. The suffixes indicate a functional relation of the nu- 
clear meaning: 'location of x," 'agent of x/ 'existence of x," etc. The case endings 
of nouns and the personal endings of the verb finally relate these words to 
other words in the sentence. This should not, however, be understood as an histo- 
rical statement in the spirit of the 19th century linguistics when it was for some 
time believed that people origmally used m their speech naked roots to which 
rootlike elements were attached until the latter became gradually mere suffixes 
to the root. Panini"s rules are purely descriptive and synchronic; the furthest 
concession ever made by Sanskrit grammarians to an historical approach is the 
recognition of uirudlui'-lahscii^ stabilization of an originally secondary mean- 
ing. Panini s suffixes are as much abstractions as his roots. When he derives 
from utsa- 'well" an adjective autsa^ 'being in or produced by a well" by adding 
to utsa- a sirifix the only visible effect is the vrddU in the first syllable: 
Msa-¥a^>auUa- (IV 1 86), besides the capacity to be inflected in any of the 
tlmee genders as demanded by its noun of reference. Prom the (abstract) root 
dis point" Panim derives the (inflectahle) noun dis- fern, 'direction’ by adding 
a suffix *^VI« which disappears after creating the accented noun: dis + ^Yl^> 
did (UOL 2 59). These 'zero suffixes"— rediscovered for modern linguistics by 
P. de Saussure— may have created the frame of mind that led to another great 
achievement in India: the position value of ciphers including zero.^a 
Prom correspondences like patati papdta; vidyd veda vaidiha; huddJia bodha 
bauddJui; kHi karfy karya a scheme of vowel alternation (ablaut) was abstracted 
m which the first and shortest vowel constituted the unnamed base form, the 

96 Ve^e quoted in theMahabhasyal 75, 13f, andIII251, 12f. Seebelowp. 124f. 
96 W.S, AiiEir, IL 16.106-113; S. Al-Geobghb, E&W 17. 115-124. 



second the ‘quality’ (guna^^) form, the third the ‘growth’ (vrddhi) form. Again 
this is not an historical statement on the relative age of the ablaut vowels. Before 
certain suffixes the base vowel of the root (or of another, preceding suffix) is 
replaced by the guna or vrddhi vowel.^^o The roots are usually^®! given in the 
base form because this constitutes the smallest unit that still contains the full 
meaning. The beauty of the three-level scheme is marred by two irregularities. 
Panini had not recognized that parallel to i e ai and uo an the base form cor- 
responding to guna /a/^o^ and vrddhi /a/ is ‘zero’. Panini explains the reduplicated 
aorist a-pa-'pt-am (from the root pat ‘faU’), therefore, as an a-aorist with an ad 
hoc invented infix /p/: a-palpit-am (VII 4 19). The roots following this ablaut 
type are taught with the guna vowel: pat, pac, etc. One could argue that these 
roots would be hard to pronounce with a zero vowel; if they could be, we would 
still need to know where to put the guna or vrddhi vowel. It is possible that 
Panini for these reasons chose to teach these roots as he did; but with Ms usual 
ingenuity he could certainly have found a solution to this problem and it 
remains probable that he had not recognized the fuU ablaut scheme. Whether 
Panini failed to recognize a second type of ablaut (rrard;i ya yd; u va vd) is 
hard to say. He teaches roots with such ablaut in the gwm grade: prach,^^^ vac, 
yaj, etc. and institutes a special procedure to acMeve the shortest vowel grade, 
the ‘stretching’ (sarnprasdrana) of the semivowel mto the corresponding vowel 
and consequent loss of /a/: prch, uc, ij. Had Panini given these roots in their 
shortest form, this would have called for a special marker or a list in order to 
prevent the usual gunxi and vrddhi substitutions {^^parch, **e/, ^*pdrch, 
etc.). This would not seem overly comphcated and it appears likely, therefore, 
that Panini had not recognized this type of ablaut as an independent variety. 

Whenever grammar is developed as an ancilla philosopMae there is a danger 
that logical or metaphysical categories are forced on the grammatical analysis: 
subject, predicate, substantive, etc. In India grammatical analysis preceded, 
due to the role of Vedic tradition and the techniques of text preservation, the 
systematic philosopMes. Panini gives us neither a logical nor a psychological 
but a grammatical description of his language. Words that end in one of the 
case suffixes 5 " . . . are our ‘nouns,’ those that end in personal endings from 

99 Lit. ‘-fold’ (L. Kenotj, JA 233. 139-142). 

100 These vowels may be further affected by internal sandhi: cikai-a > cikdya. 

101 Exceptions are roots of which no forms with the base vowel commonly 
occurred in the language. 

102 Eor reasons given below, guna Iq>I is never ruled in and rarely alluded to (ef, VI 

I 97). The main use of a in 1 1 2 is for the gwna of /r/, i.e. /ar/ (with the assistance of 

I I 51); cf. below, p. 163, fn. 7. 

103 This is one instance where internal and comparative/historical reconstruction 
differ: Panini separates gacchati {gam> gach; VII 3 77) from prcchati. The I.E. type 
g^rp-sho prlc-sho was no longer transparent in Panini’s tune due to phonological 
developments, prach offered a base for the derivation of hoth. prcchati (Vf 1 16-1-73) 
and pradna (HI 3 90; VI 4 19). 


H. Scharfe • Grammatical Literature 

tip to wwfti" our Verbs.’iM The verb is used in the second person if the word 
‘you’ stands with it in congruence or at least could stand with it; m the first 
person if ‘I/we’ could go with it; and in the third person in the remaining cases 
(1 4 105-108). The verb [-ending] denotes in itself the agent (in the active and 
voice) or the object or existence [of the action] (in the so-called passive). 
The action expressed by the verb has a large potential to supplement itself with 
expressionsfor ‘factors’ (karaka) of the action such as instrument, location, agent, 
object, etc. which can be expressed by noun cases or by other suffixes. The 
concept of kdraka-s helps to separate and link at the same time logical and 
grammatical notions. 

The algebraic fomnlationios of Panini’s rules was not appreciated by the 
first Western students; they regarded the work as abstruse or artificial. This 
criticism was evidently not shared by most Indian grammarians because several 
of them tried to outdo him in conciseness by 'trimming the last fat’ from the 
great teacher’s formulations: notably Devanandin, the author of the Jainendra 
grammar, and Vopadeva, the author of the Mugdhabodha. The Western cri- 
tique was muted and eventually turned into praise when modern schools of 
linguistics developed sophisticated notation systems of their own. Gra mm ars 
that derive words and sentences from basic elements by a string of rules^o® are 
obviously in greater need of a symbolic code than paradigmatic or direct- 
method practical grammars. 

It is interesting to observe in contrast what Panini does not teach. First he 
does not deal with phonetics. When he appears to do so, the thrust of his rule is 
nevertheless grammatical: terming sounds with an equal pronunciation effort 
in the mouthio? sa-mrrm 'of equal colour’ (1 1 9) is necessary for the use of this 
term in several later rules (e.g. I 1 69; VI 1 101). The term anundsika 'nasal 
sound’ and the three pitch accents were possibly defined because they form 
part of the metalanguage.ios The reason for this exclusion of phonetics is that 
they are presupposed, because "grammar is the science studied later on” 
(vyakaramm ncim^yam uttard vidyd, Mahabhasya 1 208, 19). 

104 The terms ndman ‘noun’ and dhhydta ‘verb’ were known to Panini because he 
teaches in IV 3 72 the formation of ndmiha and dkhydtika (or ndmdkhydtilca^ which 
obviously refers to works or a work dealing with nouns and verbs. 

10® Often mislabeled in the past as ‘nmemotechnical devices.’ 

106 This string can be very long. The fixation of the root vowel (i. e. r > ar) alone 
in the aorist form ajdgarisam ‘I woke up’ involves the application of nine sutra-s 
(VI 1 77; Vn 2 1 through 5 and 7; VTI 3 84-85) in an amazing zigzag of rules and 
restrictions. Important is the reduction to a few basic and recurrent processes. 
Processes that apply in different stages of the word formation are stated only once, 
e.g. the replacement of simple vowels by the corresponding semi -vowels in internal 
and external sandhi (VI 1 77). 

10’ P. Thieme, GGA 1958.42. G. Cabdona, Lg 41.226, proposes ‘homogeneous’ 
which is an awkward term when applied to the diphthongs. 

10® G.Caedona, Pratidanam [Fs. F.B. J. Kuxper] (The Hague, 1968), p. 458-461 ; 
H. SCHABFE, Panini’s metalanguage, p. 39. Differently, P. Thieihe, Studies [Fs. 
J. Whatmough] (’s-Gravenhage, 1957), p. 265-267. 



While Pamni teaches the formation of many nouns he is not interested in 
their lexical aspect, nor does he teach grammatical gender. The root list like- 
wise is free of lexical meanings. Due to the relatively free word order in Sanskrit 
there are but few references to such^ 

It is a puzzling questioniio why Panini has failed to derive some common 
Sansk rit nou ns with l ucid root and ablaut relations: Tnanas from ]/ Tnan, sravas 
from ]/ gVz4, c etas from i/cit Similarly havis from ]/^and yajus from ]/^, chatira 
from ^chad and saUu from -^sanj are not taught in his grammar although the 
suffixes are occasionally alluded to— they must be w^-adi suffixes (cf. our UMdi- 
sutra no. 628-678). L. Eenoui^i has suggested that maybe these formations 
were not taught because they had ceased to be productive. But Pamni has not 
hesitated otherwise to derive archaic nouns (in his Vedic rules) and he has 
even taught a suffix for the benefit of a single noun: he derives gathaha ‘singer’ 
from ^gai with the unique suffix thaka^^ (III 1 146 ).ii 2 

Though it is sometimes said that Panini’s grammar lacks instruction in 
syntax, we have seen above that this is not quite correct. Syntax is amalgamated 
with the formation of words in the process from thought to its verbal expres- 
sion, I find it surprising though that on the one hand the dominant role of the 
verb (and some verbal nouns called kHya, etc., Ill 4 70f.) is fully appreciated, 
and yet on the other hand, no attention is given to other frequent nominal 
(i. e. verbless) sentences of Sanskrit. Katyayana felt the shortcoming and sought 
to remedy it. One could amend the rule for the nominative case ending so that 
it covers also instances of identification: vlTcih 'puTusah ‘the man [is] a hero’; 
the amendment is however found unnecessary because the identification is 
obtained from the sentence as a whole. Or one could assume that the word 
‘is/are’ must be understood in all ‘verbless’ sentences.ii^ The lack of any refer- 
ence to this type of nominal sentences is not quite consistent with the usual 
process of sentence and word formation in which, as a rule, the options are 
spread out for selection. 

Panini’s description is so complex that the author must have reworked it 
repeatedly, and one can only wonderif this formidable task couldhave been carried 
out without written notes. Writing was certainly known to Panini as he himself 
refers once to script {lipillibi ‘script’ m 2 21). Whether Panini used written 
notes or not, after the task was completed the work was handed down by oral 
tradition like other texts of that time. The result is a work that is not easy to 

J.F. Staal, Word order in Sanskrit and universal grammar (Dordrecht, 1967). 
It is also curious that Paniai has failed to teach the formation of ekadasa{n) 
‘eleven(th)’ when he teaches the parallel dvddasan ‘12’ and astadasan ‘18’ (VI 3 47). 

m L. Renoxj, JA 244.159, fn. 2; cf. also G.B. Palsulb, JUP 27.145-151, who 
shows the weakness of all explanations offered so far. 

112 WACKERNAGEL-DEBBiminEB, Altuidische Grammatik, vol. II, pt. 2, p. 536 
§367ba andp. 722 §536. 

113 Varttika 1 on II 3 46 (cf. P. Thteme, JAOS 76. 3 f.) and varttika 11 on II 3 1. 


H. Scharfe * Graniniatical Literature 

nnderstand, every syllable frougbt witb meaning, often a double or triple 
meaning i even the se(pience of the rules is often important. At the same time, 
one cannot understand a rule properly if one lacks instant recall of every rule in 
the Astadhyayi as only careful memorizing can give it. Through Panini’s 
associative digressions many items are treated in unlikely sections and whoever 
attempts to follow through the formation of a word has to jump forward and 
backward in the grammar from one rule to another. This has earned the 
Astadhyayi the nickname of the ‘untimely grammar’ (a-lcdlaham vyakaranam, 
Candraviiiti on 11 2 68).ii4 The first explanation of this expression is found in 
the Nyasa on Ka^ika II 4 21, viz., that Panini’s grammar has no section on 
grammatical time, and a few centuries later the Padamahjari, on the same pas- 
sage, elaborates that in contrast to previous grammars Panini’s Astadhyayi is 
devoid of metarules on ‘today’s,’ etc. In the 14th c. A.D. Sayana tells us in 
his Phatuvrtti (under jud avcibodJiaTiB) that Panini was the first to break with 
the tradition. One can see how the growing distance in time lent wings to the 
commentators’ fancy; had they known any such older grammars they would 
not have failed to mention them. It must be noted that the expression is first 
found in a Buddhist text, the Candravrtti, from which the Ka^a frequently 
horrows,ii5 and that a-hdlaha is a Buddhist vinaya term found in the Sanskrit 
Mahaparimrvana-sutra 40.54, Divyavadana 130.22 and Mahavastu I 306, 13 
meaning ‘provided at odd times.’ It voices a critique on Panini (the use of the 
Buddhist term must have had an irouical effect) from a scholar who tried to 
improve on Panini’s grammar - even if Candragomin should not be the author 
of the honmot. The arrangement of rules in the AstadhyayP^® has indeed called 
for more attempts to improve it than an 3 rbhing else and it is not hard to see 
why later a work like the Siddhantakaumudi was so eminently successful. 

The first European contacts with Panim’s work in the 18th century were 
indirect, through popular handbooks based on Panini’s analysis until soon 
afterwards Henry Th. Colebrooke mastered the original text and translated it.i^'^ 
But only Otto v. Bohtlmgk gained for Panini his rightful place in the centre of 
Indological studies; after his first edition of 1839/40, it was his second edition 
of 1887 with translation, notes and detailed indices that has been the mam 
tool of research in this field. While his translation has been variously improved 

114 The material has been conveniently compiled by S.D. LAnntr, IL 25,187- 
199 (also in XXVIth International Congress of Orientalists, voL III, pt. 1, p. 99- 

F. Kielhobk, IA 15.183-185. 

B. Fadbegok, Studies on Panini’s grammar (Amsterdam, 1936), p. 49-68; 
M.Fowleb, JAOS 85.44^47; J.F. Staai., JAOS 86.206-209. Already Katyayana 
VI 4 1 vartt. 5-11 criticizes the creation of the a%a-section (VT 4 1— VII 4 97) 
that tears many rules from related rules taught earlier. 

Unpublished manuscripts in the Gottingen University Library; only about 
one quarter of the rules (mostly dealing with accents and Vedic forms) remained 
untranslated (F. Kielhobn', GGK 1891. 101-112). 



since, the important indices are still unrivaled.ns In a period rich in acrimonious 
debates, Theodor Goldstiicker challenged many established doctrines and their 
proponents in his Tanini: his place in Sanskrit literature’ (1861); the elegance 
of his style has gained some of his ideas a longer life than they deserve, espe- 
cially in India where his arguments are still being refuted. Even though Franz 
Kielhorn’s work centered around the Mahabhasya he furthered the study of 
Panioi’s work as well by investigating the ways in which Panini’s sutra-s have 
been amended and supplemented by later grammarians, and by clarifying the 
inner workings of the grammar. 

Acquaintance with the Paninian analysis of root and suffixes and his recogni- 
tion of ablaut— though only indirect via Ch. Wilkins’ Sanskrit Grammar- 
inspired Franz Bopp and others to develop the imposing structure of Indo- 
European comparative and historical linguistics. The generality of phonetic and 
morphophonemic rules was rigidly established only in the last decades of the 
19th cent. ; at about the same tune the notion of ^becoming’ gave way to that 
of ‘substitution.’ A purely grammatical description of language and a formalized 
set of derivational strings are hotly debated issues today. It is a sad observa- 
tion that we did not learn more from Panini than we did, that we recognized the 
value and the spirit of his ‘artificial’ and ‘abstruse’ formulations only when we 
had independently constructed comparable systems. The Indian New Logic 
(Tiavya nydya) had the same fate: only after Western mathematicians had 
developed a formal logic of their own and after this knowledge had reached a 
few Indologists, did the attitude towards the navya nydya school change from 
ridicule to respect. A striking example of how we only understand what we 
already know is the frequent translation of varim as ‘letter’ by F. Kielhom and 
others who followed the Western grammatical tradition at least in their choice 
of words, while the linguistically inclined 0. v. Bothlingk at the same time 
correctly used ‘Laut’ (e. g. in his translation of 1 3 9 and in the index under varvu). 

The last decades have seen a revival of Paninian studies both in India and 
the West (notably in the U.S.A.). This stretches from antiquarian interest^^o to 

Additions and corrections by B. Rocher, in Kjratylos 10. 69f. Before accented 
Sanskrit texts were known, 0. v. Bohtlingb: Mem. Imp. Akad. (St. Petersburg VI, 
6, Iff., “Ein erster Versuch uber den Akzent im Sanskrit”), described the Sanskrit 
accent on the basis of the grammatical treatises. B. Delbruck, IF (Anzeiger) 
17.132f. recalls the shock Bohtlingk felt when shortly afterwards accented texts 
became available that seemed to contradict his deductions — ^until their notation 
system was understood. See also L. Renou, La grammaire de Panini, Paris 
1948-1954 (La French transl. and notes based on Indian commentaries). 

1^9 E. Obermiller’s attempts in the 1920’s to write a Russian grammar in 
Paninian style (cf. Th. Stcherbatsky, IHQ 12. 380), if any drafts of it could still be 
found, would make interesting reading. 

120 After what has been said, it is not advisable to rely for antiquarian studies on 
materials supplied solely by the Ganapatha. Unfortunately, this is done frequently. 
Because of the nature of the sutra text, there can be in it only the shortest of refe- 
rences, mostly a bare noun (name): Vasudeva and Arjuna (IV 3 98), Bhdrata and 
Mahdhharata (VI 2 38). Beyond that we are dangerously free to speculate. 


H. Scharfe * Grammatical Literature 

studies on Ms grammatical theory and method of description. The problem in 
studying Panini’s method has often been a premature identification with one’s 
own theories ; we have first to find out what Panini’s conceptions are before we 
can use them to support our own. The attempt of Indian scholars to improve 
our understanding of the Egveda through Panini’s rules has not yielded the 
hoped for results, while the comparison of Panini’s language with the Middle 
Indo-Aryan languages^^a has not been pursued vigorously. 

196^!?. “ linguistics, edited by Th. Sebeok, vol. V (The 

122 F.Kielhoen, juAS 1898. 20f.; 0. I^ahee, BB 16.64r-120. 

Chapter III 


The ancient science of niruhta 'etymology’ is for ns more or less synonymom 
with the work of Yaska. The Nimkta is a commentary on the Nighantu, a 
Vedic glossary in five chapters. The first three chapters give gronps of synonyms 
in the way of the later hosa-s (mostly nonns, but there are also a few gronps of 
verbs given in the 3rd person singnlar),! The fonrth chapter has three large groups 
of rare forms and of homonyms and the fifth contains classes of divine names. 
Ko author is named for this glossary.^ 

After a lengthy introduction (1 1 to II 4), Yaska offers a running commentary: 
in II 5 to III finis he comments on the synonym lists of Nighantu I-III, though 
not on every word given there; the three batches of odd forms from Nighantu 
IV are treated in chapters IV to VI, and the six classes of divine names (Nighantu 
V), in chapters VII to XII, preceded by a lengthy discussion on theology at 
the beginning of chapter VII.^ 

The text of the Nirukta has come down to us in a shorter and a longer 
version; the word-for-word commentary^ of Durgasimha (c. 13th cent. A.D.), 
written in a Jammu hermitage, represents a third still shorter version. A study 
of the versions shows that the text grew through many small insertions and 
a new chapter of Addenda {'pcLvisistcL i later split into the two chapters X III and 

1 The first of these chapters deals with the physical world, the second with man 
and the third with abstractions. 

2 On the basis of Nirukta VII 13 samdmane ‘I enlist . . B. Bhattacharya 
[Yaska’s Nirukta (Calcutta, 1958), p. 31-33] assumes that Yaska is also the author 
of the Nighantu; but Nirukta 1 1 states: 

“A traditional list has been handed down ; it is to be explained. 

3 The first half of the Nirukta (chapters I to VI) is called naigamam handam in 
the colophone; the latter (chapters VII to XII), daivatam Tcdridam, 

^ It is curious that Durgasiipha refused to explain the Bgveda stanza III 63, 2 
quoted in Nirukta IV 14: “The stanza in which this word (i.e. lodham) occurs is 
hostile to Vasi§tha and I am a descendant of Vasi§tha, belonging to the Kapis^hala 
branch; hence I do not explain the stanza” yasmin nigama esa sabdah sa VasistJia- 
dvesinl rh. aliam ca Kdpisthalo Vdsisthah, atas tarn na nirhravlmi. The oldest extant 
coinmentary is that of Skandasvamin. K. Ktjhjtjnni Baja, ALB 28.250-26-,, 
reports on a fragment of Nilakantha’s Nirukta-^okavarttika written m Kerala 
(14th c. or earlier). The Nirukta-bha?ya of Ugrabhuti or Ugracarya (18th c.) has 
not yet been edited. 


H. Scharfe • Grammatical Literature 

XIV) still uuLnown to Durgasmiha.^ Even the text commented on hy Durga- 
simha contains insertions and this author frequently mentions variant read- 
ings. ^ 

The question of Yaska’s date has so far not been settled. All that can be said 
■with safety is that he is older than Patahjali and the flokavarttika-s quoted by 
him in Ms discussion of Pan III 3 1 (Mahabhasya II 138, 3-'19). Yaska is also 
quoted repeatedly in the Brhaddevata.'^ But the crucial question is his relation 
to Panini. The first impression of Western scholars was that Yaska must be 
older because his outlook is ritualistic and his treatment of language primitive 
•when compared with that of Panini. But these arguments are not conclusive: 
theories in parallel sciences need not develop in lockstep. Panini’s knowledge 
of a name Yasha (Pan 11 4 63) proves nothing. 

P. Thiemes has pointed out that Yaska, when he refers to regular formations, 
frequently betrays a familiarity with Panioian technique and terminology. He 
does know the concept of the root because he explains the gerund gatvd ‘having 
gone’ as an example of the loss of a final sound [[gam > gd\ + tvd, II 1) ; and he is 
aware of the suffix classes called hrt and taddhita (primary and secondary suf- 
fixes, II 2). When Yaska mentions the loss of the initial vowel of the root yas 
be in the 7iivTttisthd7ia^s (II 1) he speaks the language of the ancient Paniniya-s : 
the basic injunction to substitute a gum vowel is ‘turned away’ (nivrtti) before 
certain suffixes, i.e. we have the weakest grade — ^Yaska’s expression presup- 
poses a rule like Panini VII 3 84 being ‘turned away’ by a rule hke Panini 115: 
s-'talv^ s-UTitt, 

While this is no proof that Yaska refers to Panini’s work rather than to that 
of a forerunner of Panini, it shows that the apparent archaism of the Nirukta 
is not Papim in the making,”® but the special (in some ways conservative) 
position of the etymologists. In several instances the wording of the Nirukta 
and the Astadhyayi is very close or even identical: the definition parah samni- 
kar^ah sanJiitd “The closest conjimction is [called] connected [speech]” is 

5 In some manuscripts of the commentary there is an additional section com- 
meriting on the parihsp.. 

introduction to his commentary on the Nighantu 
If describes the poor state of the tradition 

of that text [L. Sabup, The Nighantu and the Nirukta (Lahore, 1927), pt. II, p. 

T®* are not found in our Nirukta (B. Bbiatta- 

I'-’ f'rukta, Calcutta, 1958, p. 46-56). When Katyayana (varttika 

etvmolomh^r)? the transmutation of sounds (varna-vyatyaya) as an 

Mif summarized Nirukta H 1 [P. Thibme, pfnini and 

5 L P- the other hand, S.D. Laddtt (VIJ 

s ^ y°’“Ser than Katyayana. 

P. TmEME, ZDMG 89.»23*f.; GGA 212.46-48. Cf. also M A Mbhbni>at,b 
linguistics (Bombay, 1968), p. 1-14, and G. Caedoita! 

® S.K. Belvaekae, Systems of Sanskrit grammar (Poona, 1915), p. 5. 



found both in the Nirukta I 17 and Pardni I 4 109. Twice Yaska uses a deter- 
minative: in XI 24 the prefix d is technically called a" as in Panini I 3 20^0 and 
in II 2 he derives haksa ‘armpit’ from the root gdh with a suffix Though 
Panini nowhere teaches such a suffix, he refers in VII 2 9 to an u^-adi (?) suffix 
sa (as in vat-sa ‘yearling, calf’). It means little that our Un-M-sutra 342 derives 
Icaksa from the root has with a suffix sa because these sutra-s are of a much later 
period. While we cannot be certain that Yaska knew Panini, he must have 
known a grammar so close to the Astadhyayi as to be almost identical with it. 
Considering that Panini lacks familiarity with the White Yajurveda (studied 
m the more eastern parts of India) while Yaska quotes from all branches of the 
Yajurveda, it is not hard to assume that Panini preceded Yaska and did not 
know his work. 

The most interesting part of the Nirukta is for us the detailed introduction 
in which Yaska defines his aims and methods. In the course of these discussions 
he gives us more information on early grammatical studies than any other 
author. This is all the more valuable as two of these earlier scholars, Sakatayana 
and Gargya, are also mentioned in the Astadhyayi, showing they definitely pre- 
ceded both Pani n i and Yaska. Almost^^ aU other information on pre-Paninian 
grammarians in later literature is suspect. 

Set against Panini’s formal-grammatical attitude, Yaska’s interest in philo- 
sophy is remarkable. He gives in Nirukta 1 1 a possibly traditional classification 
of words: noun, verb, prefix and particle ; the first two are established by defini- 
tions, the remaining by enumeration. Though the noun is named first, the verb is 
evidently more important and is dealt with before the noun .12 The verb has 
‘becoming’ (hJidva) as its basic notion, the noun has ‘existing thing’ (sattva);^^ 
if there should be two expressions for ‘becoming’ the one with a tune sequence 
will be expressed by a verb, the consolidated whole by a noun {pacati ‘cooks’ vs. 
pakti ‘cooking’). Then Yaska suggests a formal characterization: ‘"ados ‘that’ 
is a reference to existing things: cow, horse, man, elephant; hhavati ‘becomes’ 
[is a reference] to becoming: sits, sleeps, goes, stands. This anticipates Patan- 
jali’s statement that a verb denotes ‘action’ because it is in potential congruence 
with the verb ‘to do’: kirn karoti? pacati “What does he do? He cooks;” or 
that it denotes ‘being’ because of its congruence with the verb ‘to be’: hJimati 

10 Yaska derives the word dgas ‘sin’ from -^gam with the prefix di ‘that which 

Patanjali and a verse quoted by him (Mahabha^ya H 281.3-5) have some 
apparently authentic information on a rule of Api^ali. 

ndman ‘noun’ precedes of necessity dhhydta ‘verb’in the compowoidndmdJchydte 
‘noun and verb’ because it has less syllables. 

13 J.A.B. VAN Buitbnen, JAOS 77.104; cf. Vakyapadiya III 1 35 where hriyd 
and sattva are characterized as having and lacking sequence, respectively. 

1 ^ ada iti aattvdndm upadeiah\ gaur aivdh puruqo hastiti. bhavatiti hhdvasya: 
date iete vrajati ti^tkatUi (Nirukta 1 1). 


H. Soharfe • Grammatical Literature 

paxMti “It is [that] he cooks.”i5 Yaska’s association of the noun with sattva and 
the verb with hJiava has been quoted in the Rgveda-prati^akhya XII 5 (18-19), 
Brhaddevata 11 121, Arthasastra 11 10, 16 and Vakyapadiya II 342; often we 
iSnd dravya 'thing’ instead of sattva, and hriyd 'action’ instead of hhdva, e.g. 
Mahabh^ya II 418, 14^16, Brhaddevata I 44/45, Arthasastra II 10, 17 and 
Vakyapadiya 11 342. 

In contrast it is remarkable that Panini has kept philosophical notions out 
of his grammatical description; exceptions are only apparent. While the adjec- 
tives sioha 'little,’ aljpa ‘small,’ etc. often denote an 'existing thing’ (sattva; e.g. 
stokena visena 'with a little poison’),^® there are instances when they do not; the 
ablative ending may be used to express the semantic notion 'instrument:’ 
instrumental stokena or ablative stokdt 'hardly, with difficulty’ (Panini II 3 33 
. . . a-sattva-vacanasya). The interest of this problem reaches far beyond gram- 
mar. A certain similarity of some Indian philosophical systems (notably Nyaya 
and Vaisesika) with ancient Greek thought has struck many observers, espe- 
cially when these were contrasted with Chinese, etc. philosophies. Historical 
contact between India and the Hellenistic world can explain at best only single 
features — and the skepticism of scholars regarding such borrowing is mounting. 
The independent emergence of the categories substance, quality and action in 
the two traditions has been attributed to the common structme of their lan- 
guage: the existence of substantives, adjectives and verbs led, it is supposed, 
to the first three metaphysical categories and grammar is assigned a leading 
role in this development. Without passing judgement on the role of language 
itself,^® the role that grammatical science actually played has been reversed. 
The adjective was often not recognized as a main word class but was included 
in ndman 'noun’ as a special group the particles on the other hand, correspond 
to no metaphysical category. Philosophical categories and terms are introduced 

This notion was further developed by Bhartrhari, Vakyapadiya III 4, 3. Com- 
pare with this the remark of the American logician, W. v. O . Qunsns : “To be assumed 
as an entity is, purely and simply, to be reckoned as the value of a variable. In 
terms of the categories of traditional grammar, this amounts roughly to saying that 
to be is to be in the range of reference of pronouns” [Prom a logical point of view, 
2nd ed. (Cambridge, 1961), p. 13, quoted by B.K. Matilau, Epistemology, logic, 
and grammar in Indian philosophical analysis (The Hague, 1971), p. 110, fn. 17]. 

15 More specifically they refer to the qualities of things (gunavacana) which they 
follow in gender and number (varttika 6 on IV 1 3) ; but a quality may also be 
expressed by a noun as Patanjali points out: gavo dhinam “Cows [are his] wealth” 
(Mahabha§ya H 356. 18f.). 

17 B. Faddegok, The Vai 9 e§ika system (Amsterdam, 1918), p. 108-110; J.E. 
Staae, BSOAS 23. 109-122 and PhE&W 15, 104f. 

15 For the category ‘generality’ there is no grammatical counterpart, and note 
that action’ is exemplified in Mahabhasya I 1, 7f. by verbal nouns like ingitam, 
etc. Gf. also H. von Glasenapp, Entwicldungsstufen des indischen Denkens (Konigs- 

1® Patanjali s guTm^-sabda (Mahabhasya I 19, 20; 316, 23, etc.) and guna-vacana 
(Pamm etc.) come close to our notion of an adjective; but even then the category is 
not primarily grammatical. Cf. S.D. JoSHi, JTJP 25, 19-30. 



into grammarrso dravya ‘substance,’ ‘quaUty,’ Myd ‘action’ (besides satim 
and hhdm)— not grammatical notions into pMlosopby: ndman ‘noun,’ dkhyata 
‘verb’ and dhatu ‘root’ remain just grammatical terms. 

Regarding the prepositions, Yaska notes the controversy between Sakata- 
yana and Gargyar^i do prepositions have a meaning of their own or not? Then 
he lists 20 prepositions ( upasarga), each with one or two nouns indicating their 
value. This list (Ninikta I 3) corresponds materially if not in sequence to that 
given in the Vajasaneyi-pratisakhya VI 24, Rgveda-pratisakhya XII 6 (20), 
Paniniya Ganapatha prddayah. 

Particles are of three kinds: comparative, conjunctive, verse filler (Nirukta 
I 4) ; the latter term is secondarily expanded to ‘sentence filler’ (I 9), though a 
sentence has — contrary to a verse — ^no fixed frame that must he filled. It is 
surprising that the Indian grammarians with their astute observation have 
accepted so many meaningless fillers . 22 

Of these four word classes, a special relationship exists between the first two, 
i.e. nouns and verbs: the nouns can be derived from the verbs (or the roots 
behind them).23 Again Yaska notes the opposing standpoints of Sakatayana 
and Gargya, clearly siding with Sakatayana: all nouns can be thus derived. 
Gargya and some of the grammarians derived only those nouns where accent, 
formation and derivational modification^^ (viz. of the root) were regular. Gar- 
gya’s arguments are given and refuted one by one (1 1 12-14). Though Yaska 
argues here the theories of Sakatayana and Gargya, it is by no means certain 
or even likely that he hterally reproduced their argumentation. 

The term used to denote word derivation is pradesa as in the following argu- 
ment of an opponent. “And it is said: the existing thing precedes becoming; the 
derivation (or: designation?) of the earher from becoming — ^which is later — ^is 
not possible [and hence nouns cannot be derived from verbs]. ”25 The derived 
form is characterized by the appropriate accent and suffix, and is accompanied 

20 The old categories of past, present and future (above, p. 80) were not grammati- 
cal but philosophical in nature. 

21 Katyayana (varttika 7 on I 3 1) sides with Gargya: prepositions modify the 
action. The controversy, whether prepositions modify the root meaning or only 
illuminate a shade of meaning already present in the root, was carried on in later 
grammatical literature; cf. Mahabha$ya III 93.21-94.1 on the question what the 
root joins with first: preposition or suffix. 

22 Cf. the Tamil grammars’ acai-nilai and acai-ccoL 

22 An exception would be the onomatopoetic words, especially bird names (Ni- 
rukta III 18); the followers of Upamanyu denied the existence of onomatopoetic 

24 vikdra. The longer version and Durgasimha in 3 out of 5 occurrences have guna 
instead (and so also Sabarasvamin on Mimamsa-sutra X 4 34) which may have been 
an explaining gloss; cf. L. Sabot, The Nighantu and the Nirukta, pt. II, p. 222 f. 
Yaska uses vikdra also for the lengthening of the vowel in rdjan : rdjd (II 1). 

25 athdpi sattva-purvo hlidva ity dhuh. aparasmad hhdvdt purvasya pradeso ndpapa- 
dyata iti (Nirukta I 13). 


H. Seharfe • Grammatical Literature 

by frdddiJca vihdra ‘derivational modification. If all these features are present 
in a fitting manner, etymology is easy. If not, one shall explain a word on the 
basis of the meaning it has in a given context without excessive regard for its 
formation; in extreme cases the similarity of a single sound or syllable is 
regarded sufficient. Never is one to abandon the effort to etymologize (II 1). 
And etymologize he does: hahsa ‘armpit’ is derived from '^gdh ‘plunge into’ 
with the suffix ^sa] or f£om^ khyd ‘make known’ with redundant reduplication: 
‘wkat is there worth seeing?’; or it is derived from ]/to ‘rub against’ (II 2). 

If the words’ meanings are uniform, their etymologies are uniform; if their 
meanings are multiform, their etymologies are multiform (II 7). This theory 
led to the assumption of an excessive number of homonyms with the possibility 
of metaphor, specialization, etc. being neglected.^? The distinction of descrip- 
tive and naming nouns was known to Yaska as the foUowmg debate shows. 
An objection to etymology was raised: every person who performs a certain 
action should be named after it; and a thing or person should be named after 
all the actions performed by it or him.^s This objection is countered by the ar- 
gument that in some cases a word is used descriptively for everybody and 
everything that fits the meaning— and then often tHs is not the case when the 
meaning has been narrowed down to a certain item. 

One of the motives for the study of etymology is that, without it, the Vedic 
verses cannot be tmderstood. For Kautsa and his followers this claim was not 
convincing: he believed the Veda-s to be free of ordinary meaning. Powerful 
magic formulae, they must be applied according to the dictates of the sacred 
texts to be effective, but their efficacy is not dependent on any meaning seen 
in their words. Besides, these words are different in many ways from those of 
ordinary speech, their sequence is unalterably fi.xed and their meanings often 
appear to be contradictory. Against Kautsa, Yaska maintains that the language 
is the same and that apparent obscurities can be explained by metaphor, 
hyperbole, etc. But Yaska misses the point when he compares the rigidity of 
the Vedic texts (a closed corpus!) with word order rules pertaining to spoken 
language. Neither Kautsa nor Yaska imply any critique of the Veda-s, but the 
dispute has occasionally been taken by modem scholars as a sign of emerging 
anti- Vedic skepticism in reaUty it only exhibits the same mechanistic-magical 
attitude towards the ritual as so many late-Vedic texts.^o 

0. Strauss, ZDMG 81.115 and L. Renou, Terminologie, p. 467, derive pra- 
d^ika from an assumed pradesa 'base;’ pradediha viJcara would then be ‘root modifi- 
„ /on. ihere is no need for such an assumption. From the attested word pradesa 

zs etymologies of Yaska (Hoshiarpur, 1953), p. 8. 

SZ If (espeeiaUy p. 180) and W. S. Alek, TPS 1948: 

m Sabttp, The Nighantu and the Nirukta, part H, p. 71 f 
0. Strauss, ZDMG 81. 119-124; P. Thieme, ZH 8 26-28. 



The bulk of the Nirukta, though of considerable interest in other respects, 
falls outside the theme of this survey. If Yaska’s etymologies are often primi- 
tive and tortured and if they lack the perspicuity of Panini’s analysis, we must 
consider that Yaska dealt with different language material: with the words 
that did not yield to Panini’s methods because they were isolated and whose 
make-up could only be explained perhaps by modem comparative and historical 

Chapter IV 


The philosophical interpretation of grammar was pursued by two authors 
whose works have not survived, Vajapyayana and Vya^. Both are mentioned 
by Katyayana in his varttika-s 35 and 45 on Panini I 2 64. Vajapyayana held 
the view that words denote the Tonn’ or universal because a word evokes a 
general picture of its objects beyond special features, just as the traditional 
rules have general value: e.g. ''Do not kill a Brahmin’’ means "nobody shall 
loll any Brahmin.” Vajapyayana guards against the assumption that the single 
object is the meaning of words because that would make general statements and 
commands impossible. But that is exactly what Vya^ proposes: all orders are 
actually carried out with individual objects, and statements such as "The dog 
died, are valid individually but not for the whole race of dogs. It is not sur- 
prising that Katyayana in summing up the arguments sides with Vajapyayana. 
The Mimamsa has always held that words denote universals because this as- 
sumption met its need for universal rules; and the Vedic scholar Katyayana 
regarded grammar as a dharma-sdsira (varttika 1, introduction). 

The contrast to the individual 'thing’ is the umversal 'attribute’ or 'quality’ 
and thus Vajapyayana’s theory was developed into a conception of syntax as 
an association of qualities;’ 'the white cow’ denotes the association (sarp^arga) 
of whiteness and cow-ness, and as both reside in one place (i. e. the cow) the 
words form a syntactic unit in conformity with grammatical rules. How much 
of these doctrines (found in works of the classical period) goes back to Vajapya- 
yana himself is not known, even if they are occasionally ascribed to him,^ 
because his work was probably long lost. While there may have been a line of 
tradition we know nothing about, these doctrines can be just logical extensions 
^ well-known position of Vajapyayana’s that words denote universals. 

We know more about Vya^; bis followers were called vyadiya-a^ or sdmgraha- 
s^nka-s;^ Ms work apparently was caUed the Samgraba^ which may be a short 
form for Samgraba-sutra. It dealt principally with the question of whether the 

3 Mahabha§ya II 284, 4. 

i969).p. 20-22. Awork on 
Panbhasa-suoana ascribed to a Vyadi [K. V. Abhyakeau, Pari- 
Paribha^,V (Poona, 1962), Introduction, p. 4f.] is spurious (, 

Shadows of some early theorists 


language sounds (including words) are permanent (nityalsiddha) or made 
(hdrya, i. e. of a passing nature).^ Patanjali’s sentence ‘‘Beautiful is Daksayana’s 
work, the Samgraha”® shows the high esteem in which the Samgraha was held, 
but there remains one uncertainty: whhe both problems, i.e. the permanence 
of words and the meaning of words, are closely related and could well have been 
the subject of one and the same work, there is no definite link between the 
passages quoted. Patanjali does not say that Vya^ was the author of the 
Samgraha nor does he say that ‘Vyadi had the patronymic Daksayana (which 
could make Vya^ a distant relative of Panini through Panini’s mother Daksi). 
The earliest extant testimony for Vya^’s authorship of the Samgraha* is 
Bhartrhari’s remark in his commentary on Mahabhasya 1 6, 22.7 If this Vya^ is 
identical with the one cited several times in the Egveda-prati^akhya (III 14 +17 * 
VI12f.; XIII15), we would get the picture of a scholar who is basically a linguist 
Indeed, his contention that words denote things refiects the attitude of every- 
day speech better than Vajapyayana’s theory and it is more compatible with 
grammatical categories. ^ The question of whether words are permanent or not— 
which was so important for the Mimamsa — was apparently left undecided after 
Vyadi weighed the pros and cons: all that mattered for him was that either way 
one must study grammar (Mahabhasya I 6, 13f.). 

From Patanjali’s remarks we can conclude that he perhaps had the Samgraha 
stfil before him and even expected his reader to be familiar with it. But all 
later references are suspect, the work having perished as Bhartrhari tells us.® 
Its stupendous size alone was remembered: it dealt with 14,000 matters!® and 
had 100,000 verses.!! Dotailed statements on Vyadi’s theoiy of language in 
Helaraja’s commentary on the Vakyapadiya can be logical extensions of his 
known theorems and the numerous quotations from the Samgraha in Bhartr- 
hari’s own vrtti on the Brahma-kanda of the Vakyapadiya can hardly have 
been taken from Vya(^’s work — also they are never linked to the name Vya^. 
The tone and level of discussion in these quotations are so similar to Bhartrhari’s 
own that it is hard to believe they could have been taken from a text older 

5 Mahabhasya I 6, 12f. and 2 If. A clear exposition of this problem and how it 
occupied the minds of grammarians and philosophers for two thousand years is 
given in Sbierishna Sarma, Jnanamulctavali [Fs. J. Nobel] (New Delhi, 1963), 
p. 182-193. 

® Mahabhasya I 468, 1 1 ; the sentence shows alternative use of the genitive or 
instrumental case denoting the author of a work. 

7 Mahabha^ya-dipika, ed. K.V. Abhyankar and V.P. Limaye, ABOM 43. 

^ B.K. Matilal, Epistemology, logic, and grammar in Indian philosophical 
analysis (The Hague, 1971), p. 107f. and p. 117. 

® Vakyapadiya II 478 . . . Sanigrahe ^stam updgate “. . . when the Samgraha had 
perished”; cf. also Punyaraja’s commentary on Vakyapadiya II 484. 

!® Mahabhasya-dipika on Mahabhasya I 6, 12, ABORI 43. *21, 4f. 

!! Helaraja{?) on Vakyapadiya II 484; Nagojibhatta’s Mahabha§ya-pradipod- 
dyota on Mahabhasya 16, 12. 


H. Scharfe • Graonmatical Literature 

than Katyayana’s varttiha-s. Bhartrhari says humbly in Vakyapadiya II 484 
that his teacher (whose name was Vasurata) produced 'this compendium of 
tradition’ (ojgarm-samgraka). It is to be expected that Bhartrhari quotes his 
teacher sometimes, and therefore I propose to see in these quotations from a 
Samgraha the tribute he pays his gum. A systematic study of these quotations 
is still wanting. 

In contrast to Vajapyayana’s concept of association, Vyadi’s is said to be a 
theory of exclusion (hheda):'^^ a word denotes a 'thing’ to the exclusion of all 
other things. Thing or substance must be taken in this context in a wider sense 
because it includes theoretical constructs such as universals, fictional objects, 
etc. An indication of this kind of thinking in an early period can be obtained 
from the dicussion in the Mahabhasya II 367, 14-23 ; because of the non-Pam- 
nian enigmatic term varti{n),^^ it is likely that Patanjah relies here on a foreign 
source. While there is no problem in attaching the suffixes 4va or 4d '-ness’ to 
words like vrhsa 'tree:’ vrJcsatva or vrhsatd 'tree-ness’ it is feared that the rule 
cannot operate for words like sulcla 'white’ to obtain sulclatva or suJclatd 'white- 
ness’ because they are not or have not varti[n). Words like vrhsa denote (prima- 
rily) a thing and (secondarily) a quality the suffix is added to the word in its 
primary meaning of a thing— which is not possible in the case of suhla ‘white.’ 
And yet, it is — ^if we assume the loss of an imaginary suffix -mat 'having . . 
then the adjective suhla 'white’ is really short for suhlamat '[a thing] having 
white’ and we can attach the suffixes 4va and 4d after the basic suhla. The un- 
derlying notion is evidently that thi n gs are associated with an implied quafity 
which can also be directly expressed with the abstract noun and treated like a 
thing; but this cannot be so in the case of adjectives: a quality cannot have a 
quality. Therefore the attempt is made to treat adjectives as nouns plus zero- 
suffix. The philosophical interpretation of grammar leads to a distortion of the 
grammatical description. 

12 Helaraja on Vakyapadiya HI 1, 5, ed. K.A.S. Iyer, p. 15, 2. 

13 L. Eenof, Terminologie grammaticale (Paris, 1957), p. 270 under varta is not 
qmte satisfactory. 

The only words for things’ without quality connotations are names of people 
one has just met (e.g. Dittha, a made-up name free of associations) : later they too 
acquire quahty connotations: DiUhatva ‘Dittha-ness.’ 

T Epistemology, logic, and grammar, p. 114-116; K.K. Raja, 

meaning (Madras, 1963), p. 191-193. It would be a different 
matter the aim were to derive adjectives from nouns, but there is no indication of 

Chapter V 


The analysis of the Vedic samhitd-^ in their respective padapdtha-s (and the 
subsequent reassembling of the samhitd-pdtha-s from the fada-patha-B) implied 
a thorough knowledge of the sandhi procedures, i. e. the ways in which isolated 
(abstracted) words interact when joined in a sentence. To attain their goal of 
perfect preservation of the sacred texts, the Veda students required also a 
sound knowledge of pronunciation techniques. This was a concern for every 
school (parisadl'parsad) or branch (sdkhd) of the Vedic tradition and hence the 
manuals devoted to this task are called pdrsada or 'prdtiidhhya. As a matter of 
procedure they start with the (historically later) 'word-for-word recitation’ and 
give rules on how to construct from it the 'text in continuous recitation:’ "The 
school-treatises of all schools are based on [the recitation of isolated] words.”^ 
A notable exception to this pattern is the tMrd chapter of the Taittiriya- 
pratisakhya which gives rules in the opposite direction, viz. how vowels that 
appear long in continuous Vedic recitation are short in the word-for-word 
recitation (and for that matter, also in ordinary language). 

The chronology of these texts has been hotly disputed for over a century and 
is not finally settled. Of at least four Pratisakhya-s (Rgveda Pr., Taittiriya Pr., 
Vajasaneyi Pr., Rktantra), one or the other has variously been acclaimed as the 
oldest. But three of these texts (Rgveda Pr., Taittiriya Pr., Rktantra) are, in the 
opinion of some scholars, the youngest of all Pratisakhya-s. The question of 
their relative chronology is so hard to decide because the Pratisakhya-s — despite 
their identical basic aim — differ greatly in design.^ This is a result of their being 
school manuals in very conservative traditions. Besides, each Pratisakhya had 
to address itself to the problems of its samhitd. For the same reasons it is so 
difficult to relate them to the works on gi^ammar and etymology. L. Renou^ 
has pointed out the great number of technical terms formed from the root grh 
{pragraha, avagraha, vdgraha^ grahaim, etc.) in the Prati4akhya-s which con- 
trasts with the extensive use of derivatives from the root kr (kartr, karman, 
kdraka, krt^ etc.)^ in grammar. We have here two very different scholastic 

1 Nirukta 117 pada-prakrtvni sarm-caraimndm parsaddni. 

2 For a recent discussion of all problems of the Pratisakhya literature see L. 
Renou, JA 248.1-40. 

3 L. Renou, JA 248.37, fn. 98. 

4 G.B. Paxsuijb, JXJP 29. 11-29 ( =PCASS Class A, No. 24). 


H. Scharfe * Grammatical Literature 

editions; and yet, occasional identical stea-s and frequent agreements in 
their expressions show that they were not isolated. 

The basic question for modern scholars has been the relation of the Prati- 
^khya-s to Pardni. The belief in the continuous progress of mankind seemed to 
have the answer: the less scientific Prati^akhya-s, each concerned only with its 
respective aamhitd in its outward form and often formulated in a clumsy style, 
had to be earlier thanPanini’s grammar with its wide outlook and concise for- 
mulation. But it soon became evident that at least some of the Pratisakhya-s 
were quite familiar with Panini’s work. Caturadhyayika I 88 mentions (without 
explanation or definition!) stems ending in -vas^ which is the form in which 
Panini teaches the suffix of the part. perf. act. -vdrrisf the same Pratisakhya 
gives in 1 87 three roots in their abstract form { sdn, man, dan) as Papini III 1 6 
does even though only one of them (viz. man) occurs in the Atharvaveda for 
which the Pratisakhya is mtended; and in II 84 it treats yusmad as the base 
form of all pronouns of the 2nd person and regards tvam te, etc. as substitutes 
(adesa) following closely Panini's procedure. The sutra-s 195-218 of the 
Ektantra-vyakarana are nearly identical with Panini’s VI 1 135-157. 

In marked contrast with these two are the Taittiriya Pr. and the Rgveda Pr. 
which strictly avoid grammatical expressions. The Taittiriya Pr. went to the 
limit; it explained forms like rahsdmsi ‘protections’ or jyotlmsi ‘lights’ by 
prescribing the insertion of a nasal after /a/, /i/ and /u/ whenever followed by 
/si/ or /si/, and then removing the verbal forms daddsi ‘you give,’ dadhasi ‘you 
put, ’ etc. as exceptions in a foUowmg rule (XVI 14 + 1 8) : all that in order to avoid 
the grammatical classification in ‘nouns’ and ‘verbs’! It is unbelievable that 
its author was ignorant of this ancient division and, therefore, his attitude re- 
veals rather a sophisticated restraint® than primitive clumsiness. His phonetic 
knowledge is greater than that found in other Prati^akhya-s, e. g., in the remark 
on the degree of nasality in nasal consonants (chapter XVII) and the remark- 
able formulation ridsikd-vivarandd dnundsikyaM “Xasal quality is given by the 
unclosmg of the nose” (H 52). ‘Continuous speech,’ which Panini I 4 109 had 
defined as ‘closest contiguity’ ('par all sarnnikarsali samhitd), is explained as 
that which is within the compass of a single breath’ eka-prdna-bhdve (VI). 

One can classify the Prati^akhya-s by the degree in which they show the 
influence of grammar: Rgveda Pr. and Taittiriya Pr., the least; Vajasaneyi Pr., 
more; Caturadhyayika, Rktantra and perhaps Atharvaveda Pr., the most. But 
does this allow any inference on their relative chronology or their relation to 
Pardm? How little the use of popular terms counts as an argument is proven 
by the occurrence of the popular expressions svara ‘vowel,’ sarndJiy-ahsara 

® For more examples see B. Lxebich:, Einfiihrung in die indische einheimische 
Sprachwissenschaft (Heidelberg, 1919), II, p. 47. 

^ ® A Pratisakhya should deal only with the formal phonetic aspects of the text ; 

fitting that chapter XHI of the Taitt. Pr. deals with retroflection 
enected inside a word. 

The Prati4akhya>B 


‘diphthong,’ ghomvat Voiced,’ etc. in a late grammatical work such as the 
Katantra. Rather than comparing intellectual levels or styles we must look for 
improvements of detail, either formal or material. 

The starting poiut for such investigations must be the Vajasaneyi Pr. because 
it has the greatest affimty to Panini. A few of its sutra-s are identical with 
Panini’s; but they allow hardly any inference about the direction of borrowmg. 
Many others are so similar that they invite comparison. The appearance is 
often that of a concise rule of Panini’s corresponding to a longer (‘loose’ or 
‘diffuse’) rule in the Vajasaneyi Pr.; but the picture changes when we look 
behind the appearance. Panini 119 defined tulydsya-frayatnam sa-varymm 
“ ‘Of the same group’ is what is of equal effort in the mouth:” thus /a/ and /a/ 
belong to the same group, and likewise /t, th, d, dh, n/ belong to one group 
because their peculiarities are caused by factors outside the mouth — duration 
in the first case; aspiration, voice and nasalization (or their absence) in the 
other. The Vajasaneyi Pr. I 43 is less brief: samdna-sthdna-karandsya-prayatndk 
sa-varnah “ ‘of the same group’ is [a sound that is] of the same place of articula- 
tion, the same organ of articulation, and the same effort in the mouth.” If this 
definition is longer, it is also more precise. One could be misled by Panini’s 
definition to call /t/ and /k/ sa-variia ‘of the same group’ because both are pro- 
duced by an occlusion, viz. a contact of the tongue, though their places of 
articulation are different: /k/ at the base of the tongue, /t/ at the alveolae. 
This concern for a misappKcation is not contrived — ^we find it voiced in the 
first varttika on Panini 1 1 9 by Katy^ayana, the oldest author of the Paninian 
school whose work we have. The investigation of other parallel rules confirms 
that the Vajasaneyi Pr. improves on Panini’s formulations but does not achieve 
or aim for the same intellectual level; after aU, this manual addresses itself to 
ordinary Veda reciters and not to scholars. If this line of reasoning needed any 
support it can be found by a study of Panini’s source material: his grammar 
ignores the language of the White Yajurveda (which includes the Vajasaneyi 
Samhita to which the Prati^akhya is an auxiliary). This is primarily due to 
geographical reasons, Panini living in the extreme northwest of India and the 
White Yajurveda at home in the Eastern provinces. Panini could hardly use 
the manual of a school he did not know, but the author of the Pratriakhya 
could draw on a scholarly work that had gained wide acceptance. 

The posterity of the Vajasaneyi Pr. to Panini is the one safe point of Prati- 
sakhya chronology. Its author Katyayana is almost certainly identical with the 
author of the varttika-s on Panini’s grammar who lived around 250 B. C.*^ If 
the other Pratisakhya-s (with the possible exception of the Rgveda Pr.) are 
later than the Vajasaneyi Pr., the whole genre is much later than earlier esti- 
mates assumed: not forerunners of grammar in a ‘Vedic age’ but works of Veda 
pandits in the age of the Maurya-s and ^unga-s. And who knows how much of 

Below p. 138. 


H. Schaife • Grammatical Literature 

the so-called Vedic literature may be late, defying the simplistic concept of a 
definite and closed Tedic period’ ? The lower limit for the Prati^akhya-s would 
be approximately the time of Patanjali (c. 150 B. C.) who quotes and explains 
in his Mahabh^ya I 207, 5-10 two sutra-s of the Taittiriya Pr. (XXII, 9-10) 
and possibly in Mahabhasya I 64, 7-9 four sutra-s of the Caturadhyayika (I 
29-32). The great number of authorities cited in the Prati^akhya-s suggests 
that many treatises have been lost, even if we may doubt that each Vedic 
‘branch’ (sdhJia) actually had its own Prati^akhya. 

The object of a Pratisakhya is stated succinctly in Caturadhyayika 1 1: “Of 
the four kmds of words — ^viz. noun, verb, preposition and particle — ^the 
qualities exhibited in the combined and in the word-for-word state are here 
made the subject of treatment. The Veda reciter had to learn how to con- 
stitute the continuous text from the word-for-word text, observing the rules of 
vowel and consonant sandhi as well as those of accentuation; that included 
also a correct pronunciation of sounds. The Pratisakhya rules correspond more 
or less to the sandhi rules in Panini’s grammar, while Panini does not deal with 
pronunciation per se. The sandhi rules of the Prati^akhya-s are narrower but 
more precise than Panini’s because they concentrate on a special text; in fact 
many rules with their references to specific verses deal with a text in its conti- 
nuous and its word-for-word form rather than with language.^ While Panini 
in his quest for generalities could legitimately employ abbreviations like 
sarvddini ^sarva, etc.,’ the Veda reciter required every item spelled out; only 
the two Pratisakhya-s of the Atharvaveda and the Rktantra make extensive 
use of abbreviated rules, obviously under the influence of grammar.^® 

The Prati^khya-s have not adopted Panini’s theory of substitution (‘Vj/ 
instead of /i/ before . . .”) but follow the popular notion of 'change’ (vikdra) 
(“/i/ becomes /y/ before. . .”); the original sound is generally given in the 
nominative case, the result of the change in the accusative, e. g. Taittiriya Pr. 
V 20 Tvakdrah sahdram . . . ^^jnj . . . becomes /^/” in accordance with the meta- 
rule tam iti vihdrah “The accusative denotes the change” Vajasaneyi Pr. I 
133.^^ The technical genitive k la Panini is found occasionally under the influence 
of grammar. The Pratisakhya-s follow Panini in the use of technical ablatives 
(‘after . . .’) and locatives (‘before . . .’). The attempt of the Vajasaneyi Pr. to 
introduce a technical instrumental for an ‘insertion’ {dgama\ in 1 137) remained 

® caiurnw/m pada-jdtdndni ndmdjchydtdpasarga-nipdtdndm samdhya-padyau gunau 

® This explains why there are hardly any optional rules in the Pratisakhya-s; the 
Vedic texts were fixed. 

10 This influence is evident in the derivation of the 2. sg. imper. edhi ‘be!’ from 
the root ias vs. ihi ‘go!’ from (Atharvaveda Pr. 68f.), which falls outside the 
proper task of a Pratisakhya. 

Cf. Rgveda Pr. I 14 (56) asdv amum iti tad-hhdvam uhtam yathdntaram “(One 
should understand that) the expression ‘this to that’ means becoming that, with 
reference to the sound which (in its relation) stands nearest to it.” 

The Prati^akhya-s 


without followers, just as Ms short terms sim ‘simple vowels,' mud ‘sibilants’, 

The main interest of the Pratisakhya-s lies for us in their descriptions of the 
contemporary pronunciation; also, they serve as a check on the tradition of 
our manuscripts of the Vedic samhitd-s. Their phonetic observations were far 
more to the point than anything acMeved in Europe before the last quarter of 
the 19th century when — ^largely under the influence of these same Pratisakhya-s 
— ^modern phonetics emerged as a science. The places of articulations (i.e. 
palate, teeth, lips, etc.), the organs (root of the tongue, tip of the tongue, etc.) 
and accompanying features were clearly separated: those inside the mouth 
(opening, closure, constriction) contrasted with those outside it (aspiration, 
voicing, nasalization) even though the role of the vocal cords was not recognized. 
And yet I hesitate to accept their phonetic description as a true image of old 
Vedic pronunciation because the mam thrust is to meet popular trends as we 
find them in the Middle Indo-Aryan (MIA) dialects .12 The Kgveda Pr. devotes 
a whole chapter (ch. XIV) to faulty pronunciations with rules such as 5 (14) 
sdntasthdndm ddi-lopdnta-lopau “In the case of those [consonants] wMch appear 
with a semivowel the first or the last [sound] is dropped” or 5 (16/17) anyonyena 
vyanjaudndrn virdgo lesena vd vacauam ^ilanam vd “Mutual colouring of conso- 
nants or partial pronunciation or suppression [of consonants].” The Middle- 
Indie somd law r>i led to uncertainty when to pronounce an original /r/: in 
the last syllable of candra-nirnih we might hear a hypercorrect /r/ instead of 
an /i/, in srnga a popular *Mnga [17 (45/46)]. 

The Pratisakhya-s devote several sutra-s to ‘vowel fractions’ (svara-bhalcti) 
wMeh are inserted between an /r/ or /!/ and a following consonant, especially 
a fricative: varsa wiU be pronounced vardsa, arhati will sound like ardhati. As 
these are the forms we commonly find in the MIA dialects (e. g. varisa, ardhanta) 
it is hkely that the Pratisakhya-s describe a popular pronunciation of Sanskrit; 
if this pronunciation should be old, we have so far no way of proving it. 

The aspirate pronunciation of stops before sibilants (e.g. vaihsa for mtsa, 
vlraphsin for vlrapsin) is taught in Egveda Pr. VI 15 (54) ; Taittiriya Pr. XIV 
12; Caturadhyayika 11 6 (and Katyayana’s varttika 3 on Panini VIII 4 48). 
This too is probably popular pronunciation.^^ In groups of stops the first is 
often incompletely pronounced: the /d/ in marudhliih is not exploded [e.g. 
Egveda Pr. VI 5 (17)] but VyaM severely restricted this rule to certain situa- 
tions [Egveda Pr. VI 12 (43/44)]. Such incomplete articulation or implosion was 
a necessary stage in the development towards the consonant assimilations 
wMch we find in the MDLA dialects. Again it is not possible to make any state- 

12 H. Jacobi, KZ 25. 603-609. 

13 J. Wackernagel, Altindische Grammatik, 2iid ed. (Gottingen, 1957), I § 
113 with Xachtrage. Cf. also below p. 144f. 

14 The same holds for stops in final position [Egveda Pr. VI 5 (18)]. 


H. Seharfe • Grammatical Literature 

ments about wben this incomplete articulation came into use. The explosion 
of the first stop in certain clusters^® is called sphotana ‘bursting’ (Caturadhya- 
yika 1 103 and 11 38). 

Besides the rules on incomplete articulation, and with no attempt to contrast 
or harmonize these doctrines, we find rules on doubling applicable in the same 
cases: the first member of a consonant group should be pronounced double 
(sapta: sappta), likewise a consonant in final position (marut: marutt). The 
earliest reference to doubling is Panini VIII 4 47 where optional doubling of 
the first consonant in a group is accepted; Panini cites Sakatayana’s opinion 
that no such doubling takes place in original groups of three or more consonants 
(Vni 4 50 ; i. e. in words like Indr ah or rdstram), Katyayana added in his varttika 
3 on VIII 4 47 the optional doubling of a final consonant. Doubling of the final 
consonant is not admitted by Vajasane}^ Pr. IV 114,^6 Rgveda Pr. VI 2 (7) and 
Taittiriya Pr. XIV 15, and was apparently limited to certain schools or areas. 
As I see it, this doubling is an attempt to speak correct academic Sanskrit. 
Against the common tendency to articulate a final stop incompletely which led 
to its disappearance in the MIA dialects, the ‘correct’ pronunciation would 
stress the explosion and give the impression of a double consonant. The same 
applies to the initial stop in a consonant cluster: the tendency towards assimila- 
tion was cheeked with a heavy articulation of the first stop. 

Comparison with Panini’s grammar reveals another peculiarity of the Prati- 
sakhya-s. Panini composed his work for oral transmission from teacher to 
pupil; it had to be memorized and aU technical implications had to be mastered 
before it could be used. When his grammar was finally put down in writing, the 
technical accents and nasalizations were lost because they could not be ex- 
pressed in the script of the time. The user of a Prati^akhya on the other hand, 
would know his Vedic samhiid by heart, but he needed no unusual mastery of 
the Prati^khya — certainly nothing that was not contained in the written text. 
The use of accents is avoided even where clarity and verbal economy would have 
demanded it: referring to the adverb antdh ‘within’ (in contrast to the noun 
dntah end ) the Vajasaneyi Pr. I 162 says antar an-ddy-uddttam ‘the antah 
that is not accented on the first syllable’ — Panini would simply have read the 
adverb with its accent. Por the authors of the Pratisakhya-s, pitch accents 
were no longer a fact of spoken Sanskrit but limited to Vedic recitation. 

The Rgveda, Pratisakhya is attributed to Saunaka; it is the only Prati^akhya 
completely made up of verses: a medley o£tristuhh-s, jagati-s, sloJca^s and a few 
other metres. It is organized in 18 chapters or patala-^ over which a mechanical 
division into three adhydya-^ (with six patala-s each) and varga-s (consisting ge- 

According to Caturadhyayika II 38 whenever a stop is followed by another of 
a more back series. 

against Katyayana’s authorship of both 
the Vajasaneyi Pr. and the varttika-s: the grammarian had to be more catholic than 
the Vajasaneyin who need not and should not teach procedures not followed in his 

The Prati4akhya-s 


nerally of five stanzas each) is superimposed. The ten introductory stanzas are 
a late addition and have not been commented on by XJvata in his commentary 
to the Prati^akhya. Closer study of the text shows that it was not cast in one 
mold. The chapters XVI to XVIII constitute virtually an independent manual 
of metrics; separate authorship has also been assumed with some hesitation 
for chapters XIII to XV, and perhaps for chapters XI and XII. The frequently 
voiced assumption, however, of old pre-Paninian kernels in this Pratisakhya 
as well as in others lacks solid proof. 

The metrical form led to a larger use of synon3rms, finite verb forms and 
connecting particles but the style is not verbose. The verses can be dissolved 
into strings of siitra-Hke sentences with frequent use of dittoing.^’ This Prati- 
^akhya has shunned arbitrarily created terms but employs more common 
words in a defined technical meaning than any other. Almost every phenomenon 
of sandhi is called by a special name which is introduced by definition and then 
never used again in the text: the loss of /r/ in abravl[r\ Rdmam is called 'a-lcdma' 
in ntfatilr] rdjate with lengthening of the preceding vowel (>nrpati rdjate) 
^niyata^ [IV 9 (30)].^^ The technical use of the accusative is taught^® but the 
abbreviated expression made possible by it is sometimes set aside in favour of 
a non-technical formulation.^o The commentator Uvata (11th c. A.D., from 
Gujerat; he wrote commentaries also on the Vajasaneyi Samhita and the 
Vajasaneyi Pr.) quotes an older com m entary (vHti) and in fact his remarks 
seem to be, with the exception of the more independent remarks on the first 
four or five 'patala-B, largely identical with an anonymous Parsada-vrtti or (in 
other manuscripts) Parsada-vyakhya. On the ten introductory stanzas we have 
a brief commentary Varga-dvaya-vrtti by Visnumitra who seems to refer to 
Uvata’s work and hence should be younger. It is remarkable how rarely Sayana 
in his commentary on the Egveda has made use of the Prati^khya. 

The Rktantra-vyakarana, the Pratisakhya of the Samaveda, is perhaps the 
youngest of the true Prati^akhya-s^i for it shows the strongest influence of 
grammar including extensive use of gaim-s {‘. . . etc.’) instead of lists. It is also 
the shortest with only 287 brief sutra-s. 

The Efrsna-yajuh-prati^akhya ‘Pratisakhya of the Black Yajurveda’ (as- 
cribed in one manuscript to Kartikeya) has been labeled the Taittidya-prati- 
^akhya by W. D. Whitney at a time when little was known of other branches of 
this Veda. Actually references to different views held by the Taittiriya-s in 

The text is quoted by papla and stanza with the sutra number eventually 
added in parentheses. 

18 Cf. Mangau Deva Sbastri’s edition, vol. I (Benares, 1959), introduction, p. 56 
and 66f.; vol. HI (Lahore, 1937), p. 322-327. 

18 I 14 (56), see above p. 130 fn. 11. 

80 II 4 (10) tatra prathamds trtlya-bhdvarn pratilome§u niyanti “In the so-called 
pratiloma anvak^ara-samdhi-s the first class consonants (i.e. tenues) become the 
third class consonants (i.e. mediae).” 

81 We can leave aside here three texts also called Pratisakhya-s of the Samaveda 
but dealing with chanting: Ak§ara-tantra, Pu§pa-sutra and Sama-tantra. 


H. Scharfe • Grammatical Literature 

XXTTI 15 and 16 led WMtney to assume that the work originated in another 
school, but the inference is weak because chapters XXII to XXIV are probably 
later additions to the text .22 Perhaps the rule XIV 9 points to Central India as 
the home of this Prati^akhya; at least the insertion of a stop between a fricative 
and a nasal (e.g. Krs\t}m, grislp]ma) is common in many later dialects of that 
area.23 The commentary on this Prati^akhya, called Tribhasyaratna, is based — 
as its name indicates — on the work of three predecessors, i. e. Vararuci, Atreya 
and Mahiseya; nothing is known about its author or his time. 

The Sukla-yajuh-prati^akhya 'Pratisakhya of the White Yajurveda’ or Kat- 
yayaniya-pratisakhya is commonly quoted as the Vajasaneyi-pratisakhya; its 
author Katyayana is almost certainly identical with the author of the Varttika-s 
on Panioi’s grammar.^Each of the eight chapters of the Pratisakhya is closed with 
the auspicious words vrddham vrddhir just as the word siddham concludes every 
last varttika in every one of the eight adhydya-s. Uvata’s commentary on this 
Prati^khya seems to be later than that on the Egveda Pr. because his explana- 
tions of certain theoretical points have become more sophisticated.^s While 
Uvata comments on the Pratisakhya from the standpoint of a student of the 
Madhyamdina branch, the commentator Anantabhatta reinterprets it as a 
manual of his own Kanva recension of the White Yajurveda. 

Of the two Pratisakhya-s of the Atharvaveda, the one connected with our 
vulgate text of this Veda, by some twist of fate, remained virtually unknown 
until a few decades ago. First a shorter and mutilated version was discovered, 
and subsequently more complete manuscripts of it became known. The 
Caturadhyayika, published firom a single manuscript more than a century ago 
by W.D. Whitney as the Atharvaveda-pratisakhya, belongs to a lost branch of 
the Atharvaveda. Because of its last colophon it has also been accepted as the 
Saunakiya Caturadhyayika even though in rule I 8 the opinion of ^aunaka is 
first mentioned and then rejected. A newly found manuscript of this text^® with 
better readings calls it instead in the colophons Atharvavede Kautsa-vydkarane 
Caturddhydyilcd ‘The [work] in four chapters in Kautsa’s grammar on the Atharva- 
veda. hard to say whether Patanjali’s sentence upasedivdn Kautsah 

Fdninim “Kautsa came to Pamni as a student’’ (Mahabhasya n 115, 17) refers 
to the same person or rather to another member of that gotra. Both Prati- 
sakhya-s of the Atharvaveda show strong influence of the techniques of gram- 

22 Of. also H. Ltoebs, Die Vyasa-Qiksha (Kiel, 1895), p. 57-59. 

o studies in the phonetic observations of Indian grammarians, 

2nd ed. (Delhi, 1961), p. 121-125. 

^ See below p. 13^141. 

T> discussion on the use of technical terms found in the commentary on 

-Kgveda Pr. H 26 and on Vajasaneyi Pr. I 54. 

Sadashiva L, Katee, NIA 1. 383-396, 

Chapter VI 


Katyayanai ^as not the first to comment on Panini's Astadhyayi but his 
vdrttika-Q [’remarks on the [teaching] procedure (vritiy (?)]2 are the first such 
work that is preserved in its entirety. Their preservation is owed to Patanjali 
who included them in his ‘great work in colloquial language’ (mahabhdsya) 
and discussed their pros and cons. We have reason to believe that no varttika 
has been left out: in the sometimes lengthy discussions (there are 59 varttika-s 
on Panioi I 2 64 and 33 on II 1 1!) we see a logical development of the ar- 
gument. When Katyayana refers about a hundred times to another varttika 
with uJctam “It has been said”^ this reference can be found (e.g. varttika 
8 on Panini 114 refers back to varttika 4 on the same sutra) ; all eight adh- 
ydya-s conclude with the auspicious word siddham “it is correct.” In isolated 
cases there may be some doubt as to whether a sentence found m the Maha- 
bhasya-text is a varttika of Katyayana or a statement of Patanjali, but in the 
overwhelming majority of cases we can recognize the varttika-s with the help 
of the criteria developed by F. Kielhom: the varttika-s are generally followed 
by a paraphrase of Patanjali; their nominal sutra-hke style differs from the 
conversational style of the Mahabhasya. Their total number is approximately 
42934 — attached to 1245 of Panini’s circa 4000 rules. There are manuscripts that 
contain the varttika-s alone but these are secondarily derived from the Maha- 

1 The differentiation between an older mahdvdrttikaJcdra Katya and a later (and 
minor) varttilcakdra Katyayana made by K.G. SuBRAMAmAM, JOB 2.25-33 and 
V. Raghavaojt, Bhoja’s Srngara Praka^a (Madras, 1963), p. 746, is not convincing. 
There is no support for the later identification of Katyayana with Vararuci. 

2 Patanjali, Mahabhasya I 371, 18, contrasts Panini*s vrtti-8utra-s ‘concise rules 
of procedure’ with Katyayana’s vdrttiJca-s ‘remarks on the procedure.’ To judge 
from the expression vdrttika-sutrika ‘student of varttika' (Mahabhasya 11 284, 3f.; 
not ‘student of varttika and sutra' MW) — if it refers to students of the same text — 
the full name would have been vdrttika-sutra ‘concise statements relating to proce- 
dure.’ Of. P. TuEEikiE, GGA 212. 23 f. 

3 Frequently this reference is to a varttika that comes later in the text, e.g. 
varttika 9 on Panini I 1 3 cdktam refers to varttika 5 on Panini VI 1 13 and 
varttika 6 on I 1 12 refers to varttika 9-4*10 on VTII 2 6. This shows that the whole 
text of the varttika-s is supposed to be present in the mind of the student. 

4 K. Mabhava Krishota Sarma, Panini Katyayana and Patanjali (Delhi, 1968), 
p. 53 f. 


H. Scharfe * Grammatical Literature 

biiasya text. The combined edition of Pardni’s sutra-s with Katyayana’s 
varttika-s5 gives an impression of the tradition as it might have appeared to 
Katyayana’s successors, but it is likewise secondarily derived. 

The careful separation of the varttika-s from the surrounding text of the 
Mahabhasya and their stylistic characterization was the work of P. Kielhorn^^ 
whose investigation generally confirmed the Indian grammatical tradition 
about the separate identity of Katyayana's varttika-s. A varttika (or the first 
of several varttika-s) on a sutra of Panioi'^ often carries as a mark of reference 
the full quotation of Panini’s sutra (131 instances) or the quotation changed 
only insofar as to allow its being construed with the other words of the varttfiia. 
Otherwise the varttika contains at least the first word(s) of a sutra or the special 
term taught in it. The absence of any such references in a few varttika-s or 
other inconsistencies in a few others raise doubts about their status, but the 
small number of such cases does not justify a general skepticism^ against the 
establishment of Katyayana as an author with a separate identity. 

More specific was E. Frauwallner’s® attempt to prove that at least in one 
instance a large group of supposed varttika-s were really inserted by Patanjah 
from another source into the argumentation of Katyayana. In his varttika-s 
on Panini 1 2 64, Katyayana discusses Panini’s peculiar way of generating dual 
and plural forms: in the case of two trees (spelled out vrhsai ca vrJcsas ca ‘a tree 
and a tree’) the dual suffix au is used according to Panini 1 4 and 1 4 103 
before this suffix only one word vrlcsa remains (Panini I 2 64^^) — vrhsau. In his 
lengthy discussion of the feasibility of this rule I 2 64, Katyayana introduces 
the opposing arguments brought forward by two other scholars, Vajapyayana 
and Vyadi (above p. 124-126) : the former held that words denote first the univer- 
sal, the latter on the contrary that they denote single things. Panini’s procedure 
would obviously imply that he took words to denote single things first — but 
probably Panini had taken no principled stand on this question which may 

5 Contained in: Word Index to Panini-sutra-patha and PariiSi§tas, compiled by 
S. Pathak and S. Cm'rRAO (Poona, 1935), p. 461-648. 

® F. KiELHonN, Katyayana and Patanjali (Bombay, 1876; several reprints), lA 
15.203-211, and above all, his edition of the Mahabhasya (Bombay, 1880-1885; 
2nd revised ed., 1892-1909). 

’’ Some varttika-s refer to several sutra-s conjointly, e.g. under I 2 29-30, I 2 
68-71, 1 4 105—108; the two varttika-s on V 4 113—115 are one sentence. 

8 Such skepticism was voiced by R. Rocher, JAOS 91.315, in whose article one 
misses a reference to Kjelhorn’s Katyayana and Patanjali. 

9 E. FRAinvAiLisrEiB, WZKSOA 4.92-106. 

^9 1 4 22 dvy-ehayor dvivamnaikavacane ‘^Referring to two or one, dual or singular, 

The second suffix in each of the triplets given in TV 1 2 is called dual. 

12 I 2 64 aa-rupanam eha-aesa eka-mbhalctau “Only one of the identical [words] 
remains if followed by the same case suffix.” 



have arisen well after his time.i^ Katyayana sides with Vajapyayana and his 
belief in the role of universals; he gives in the varttika-s 35-44 the basic argu- 
ments of Vajapyayana, in varttika-s 45-52 one by one Vyadi’s counter-argu- 
ments and finally, in varttika-s 53-59 ins own opinion which is similar to 
Vajapyayana’s: a word denotes the universal which manifests itself in indivi- 
duals (Mahabhasya I 242.10-247.16). Frauwallner, who regards the whole 
discussion on the role of universals as an insertion made by Patanjali, assumed 
that the original text excerpted by Patanjali preceded not in three but iu two 
steps: first the theory of Vya^, then Vajapyayana’s theory of universals. His 
main argument is an apparently senseless repetition of varttika 40 (in varttika 
56): asti caikam anehMhiharami-stMmyugap(^ — ittndravad visaydh “It happens 
that one thing is simultaneously in several places: the range is like [that of] 

Varttika 40 is embedded in the Mahabhasya text and its two parts are sepa- 
rated by Patahjali’s paraphrase and discussion of the first part. 

“It happens also that one thing is perceived simultaneously as being in several 
places, e.g. the sun: the one sun is seen simultaneously in several places. — This 
analogy is not correct because not one and the same viewer sees the sun simul- 
taneously as being in several places. — Then: ‘The range is like [that of ] Indra’ ...” 

This is repeated, said Frauwallner, in varttika 56 with accompanying com- 

na caikam aneJeddhikarana-stham yv^apad ity adityavad visayah “[Regarding your 
statement that] one cannot be simultaneously in several places [I say:] the range 
(of application) is like the sun.” 

Patahjah paraphrases the varttika and continues: 

“E.g. the one sun is seen simultaneously as being in several places. — The analogy 
is not correct because not one and the same viewer sees the sun as being in 
several places. — ^Then: The range is like [that of] Indra,i6 viz. the one Indra who 
is called at hundreds of rituals is at all [these places] simultaneously. Thus the 
form (universal) will be at all places simultaneously.” 

Frauwallner objected to the idea that an example (viz. the god Indra) which 
is offered in varttika 40 but replaced by another (viz. the sun) in varttika 56 

As Patanjali, Mahabhasya I 6, 8—11 points out, Panini^'s rule I 2 58 would 
imply that he saw the universals as the meaning of the word and I 2 64, conversely, 
the individuals. 

14 It must be remembered that the formulation of these varttika-s is Katyayana’s 
and that they are probably not literal quotations of Vajapyayana's and Vyadi s 

works. ..... A -rx 

13 God Indra attends many rituals at the same time. Nagojibhatta (18thc. A.D., 
in his Uddyota on varttika 40) assumes that it is the word Indra that is mvoked at 
several rituals simultaneously and becomes part of them, a late rationalistic inter- 
pretation which he tries to read into the text of the Mahabhasya. 

16 The edition of JosHi, KupixA and Raghunatha-sastrust prints the phrase 
iiindravad visayah as a separate varttika in both occurrences. 


H. Scharfe ■ Grammatical Literature 

should again be preferred in Patanjali’s commentary on both varttika-s. The 
problem disappears when we recognize that the editors erred in regarding 
iiindramd visayah as a varttika or part of a varttika.^'^ Vajapyayana’s postulate 
asti caiham aneMdhikaraTia-stham yugapad (varttika 40) was answered by 
Vya^’s counterclaim na caiham anehMhiharaTiu-stJiam yugapad (varttika 48) — 
both equally unsupported by an example. The latter is then refuted by Katya- 
yana in his definite opinion (varttika 56): 

'na caiham anehMhihararia-stham yugapad' ity adityavad vi§ayah “[Regarding 

your statement that] one cannot be simultaneously in several places [I say;] 

the range (of application) is like the sun.” 

That was the end of the argument for Katyayana just as it was for the author 
of I^Iimamsa-sutra I 1, 15 yaugapadyam adityavad “Simultaneousness like [in 
the case of] the sun.” Only Patanjali was not satisfied; if the sun is seen simul- 
taneously in several places this is due to several different viewers, which makes 
the statement relational and worthless. Instead of ity adityavad visayah “[I 
say:] the range is like the sun,” the varttika 56 should end itindravad visayah 
“[I say:] the range is Hke Indra.” This use of the particle iti (^regarding your 
statement that ... I say . . .’) is so typical of Katyayana’s style that only 
varttika 56 can be the original place of the example. With iti Katyayana refem 
back to varttika 48 just as he so often refers with iti to Panini’s sutra-s (e.g. 
on Panini 1 1 47; 1 1 52, etc.) or with iti ced ‘If you say . . . then I say , . to 
a h 3 rpothetical reasoning. Patanjali was free to mention the examples that 
were to be introduced later and no further inferences can be drawn from this 
fact. Thus, a clear xmderstanding of the Mahabhasya is only possible after a 
correct separation of Katyayana’s and Patanjah’s contributions. 

To determine when and where Katyayana lived we depend on incidental 
references. On Panini VI 3 21 sasthyd dhrose “[Before the second word of a 
compound there is non-disappearance] of the genitive ending if [the compound] 
expresses an insult” Katyayana ’s varttika 3 demands an exception — devandm- 
priyUj the title of the Maurya kings. The elliptical expression sdka~pdrthiva 
vegetable [eating] king,’ i.e. Vegetarian king’ in varttika 8 on Panini II 1 69, 
can hardly refer to anybody but Priyadarsin Asoka and suggests thus a date 
after 250 B. On the other hand, Katyayana cannot have lived much later 
than that because of the large derived literature (variant readings of the vart- 
tika-s, polemics against them, etc.) quoted by Patanjali (c. 150 B.C.) in his 

H. Jacobi, Indian Studies [Ps. Ch, R. LAinvtAisr] (Cambridge, 1929), p. 151. 
Note that itindravad visayah is not repeated or ‘paraphrased’ in Patanjali’s discus- 
sion as the authentic varttika-s are. 

In the traditional interpretation no justice is done to the use of iti, and the 
mirror image relationship between varttika-s 40 and 48 is lost. 

H. Scharfe, KZ 85.211-225, 



Scholars have long assumed that Katyayana lived in the South, i.e. the 
Dekkhan, because of a statement of Patanjali. At the end of his very first 
varttika Katyayana offers a parallel: yathd lauJcika-vaidikesu . . as in secular 
and Vedic [affairs]/’ On the question why Katyayana did not simply say: 
yathd loke vede ca , as in the worldPy life] and in the Veda,” Patanjali sug- 
gests a) that Southerners are overly fond of secondary suffixes or b) that 
perhaps there is a special meaning to the longer formulation. Patanjali’s 
suggestion has been taken by modem scholars as a statement that Katyayana 
was a Southerner but no such statement is implied; aU we may conclude is that 
Patanjali thought it possible that Katyayana was a Southerner and that this 
may explain the unusual formulation. That Katyayana indeed lived in the 
South becomes likely through his varttika on Panini VI 3 73 nalopo na^.ah 
“The negation na looses its /n/ [in the beginning of a compound]” (e.g. a-putra 
‘not a son; sonless’) which demands similar negated forms from the verb if an 
insult is intended: 

na^.o na-lope ^vaksepe ti^,y upasamkhydnam “In addition to Panini VI 3 73 it 
must also be taught that the /n/ of the negation na is dropped before a verb form 
if an insult is intended.” 

Patanjali gives the examples apacasi vai tvam jdlma “You don’t cook [right], 
you fool!” and akarosi vai tvam jdlma “You don’t do it [right], you fool!” 
Negated verb forms are strange to Sanskrit and the Indo-European languages 
in general, 20 but are a common feature of the Dravidian languages .21 And if we 
look at the few examples of such negative forms in Sanskrit texts , 22 it is striking 
that two of these texts are definitely from South India: Sahkara^s hailed from 
Kerala and the Bhagavata-purana {a-sprhayanti III 25, 37) comes from the 
Tamil country. Besides, it is noticeable that forms like a-sakkoii ‘cannot’ are 
frequent in the later Pali texts.24 Qne thing is certain: Katyayana neither be- 
longs to the West nor to the North of India because of his links with the White 
Yajurveda25 which was not represented in these areas ; nor was he an Easterner 
because in his varttika 8 on Panini VII 3 45, he postulates the bird name 
vartaka ‘quail’ for the ‘eastern’ dialect while he apparently used vartika — as does 
the Vajasaneyi Samhita XXIV 30. 

20 From the Greek, E. Schwyzeb, Griechische Grammatik (Munchen, 1939; re- 
printed, 1959), I, p. 432 and 644, fn. 3 adduces a solitary axUi ‘he does not honour’ 
(Theognis 621) which he explains as an ad hoc creation. 

21 J. Bloch, Structure grammaticale des langues dravidiennes (Paris, 1946), p. 51. 

22 L. Rbnou, Grammaire Sanscrite (Paris, 1961), p. 174. 

23 G. Thibatjt, ZDMG 48.540, has disputed the correctness of these readings in 
Sankara manuscripts. Magha’s Stiupalavadha XV 33 jdlma . . . a~gha^te shows its 
dependence on the Mahabha§ya. 

24 D. Andebsok, Pali reader, Glossajry (London, 1907), p. 2. 

25 B.A. VAK Nooten, IL 29.43-46. 


H. Scharfe * Grammatical Literature 

It is virtually certain that Katyayana the Varttika-kara is identical with 
Katyayana the author of the Vajasaneyi Pratisakhya of the White Yajurveda.^e 
We have seen above (p. 129) how the Vajasaneyi Pr. 1 43 improved on Panini’s 
definition! 1 9: 

tulydsya-prayaimrrh sa-varnam “‘Of the same group’ is what is of equal effort in 
the mouth” becomes a cumbersome samdna-sthdna-haranasya-prayatnah sa-vamah 
“‘Of the same group’ is [a sound that is] of the same place of articulation, the 
same organ of articulation and the same effort in the mouth.” 

The motive for this alteration is voiced in Katyayana’s first varttika on Panini 
119: Panini's definition is too wide as it would extend to stops such as /k/ and 
/t/ which are produced through identical (samdna) efforts though at different 
places. But in Ms solution he stays as close to Panini*s formulation as possible 
wMle removing its shortcomings. He retains the adjective tulya against sarrmta 
of the Pratisakhya and proposes in varttika 2 to read: dsye tulya-dem-prayatnarri 
sa^vatTiam ‘‘‘Of the same group’ is what is of equal location and effort in the 
mouth.” The varttika-s indicate familiarity with the Vajasaneyi Pr. I 43 and 
must therefore be later; the solution proposed by the varttika is superior. 

Panini 119 tvlydsya-prayatnam sa-varTUtm: concise but not precise 
Vajasaneyi Pr. 1 43 saindim-stJidna-harandsya~pray sa-varyxih : not con- 
cise but precise 

varttika 2 on 1 1 9 dsye tulya-desa-prayatnam sa-varnam: both concise and 

In the writing of his varttika-s, Katyayana has not been able to free himself 
completely from the terminology and the attitudes of the Prati^akhya-s. He 
uses^occasionaUy svara for ‘vowel’s? instead of s'parsdghosa for ‘voiceless 
stop 28 instead of etc. ; we even find svastani ‘2nd future suffix’29 instead 
of LuP and bMvardl ‘present tense snffix’^o instead of Lah Instead of Paninian 
substitution (ddesa) of sounds he speaks occasionally of ‘change’ (vilcdra)^^ 

9i9 L.Renotj, JA 230.169-176; P. Thieme, GGA 

bABMA, Panim Katyayana and Patanjali, p. 108f., raises the 
objection that according to Vajasaneyi Pr. I 73 the two components of the diph- 
th^gs /ai/ and /au/ are equally one measure (mdtra) long, while according to 
varttika 4 on I I 48 the Mphthongs are made up from 1/2 measure /a/ and 11/2 
meastues /i/ and /u/. This interpretation reads more into the texts than they intend 
to state, me Pratisakhya merely speaks of the measure of /a/ being guttural, that 
1 respectively. The varttika only says that the latter 

part ot the diphthong is greater, which Patanjali, Mahabha^a I 118, 2f., explains 
by saymg that the measure of /i/ and /u/ is greater, the [measure] of /a/ lighter;” 
the term measure’ is not used here with the precise time value. 

^’E.g, varttika 5 on I 1 7. 

28 E.g. varttika 7 on I 4 109. 

2® Varttika 1 on III 3 15. 

20 Varttika 11 on H 3 1. 

2^ E.g. varttika 15 on Siva-sutra 5. 



and even uses a teclinical accusative to denote such a ^change’^^ instead of 
Pamni’s genitive of substitution. But Katyayana’s obligation to Pratisakhya 
techniques goes still deeper and touches on the basic difference between gram- 
mar and Pratisakhya. Grammar strives for scientific generalization, for the 
essence of things; the Pratisakhya-s look for practical rules to aid the priestly 
practitioner, with every detail spelled out. 

Panini’s rules I 2 37-38 give rise to a discussion in which Katyayana displays 
his dual approach. The subject is a Vedic mantra of over twenty words, the 
so-called Subrahmanyd.^^ Panini had succinctly defined the peculiar way in 
which it is recited: aU vowels with (normally) falling tone are chanted in high 
pitch with the exception of the two words devah and brdhmdTmh at the end of 
the formula where these vowels are chanted in a low pitch instead. Panini’s 
description is marred by a technical oversight: several of the falling tones are 
only secondarily established in one of the very last rules of his grammar (VIII 
4 66^), in that final section of strictly linear rules called the tripddl — whence 
these falling tones are not available for substitution in I 2 37-38. All would he 
faultless if the rules I 2 37-38 would he put after VIII 4 66; hut — and that is 
Katyayana’s final opinion in his varttika 5 on 1 2 32 — ^such transfer of the rules 
I 2 37-38 is not necessary because Panini’s procedure indicates that he himself 
regarded the rules I 2 37-38 as if they followed VIII 4 66. Katyayana argues 
that it would make no sense otherwise to say that the falling tones of devdh and 
brdhmanah are chanted not in the high but in the low pitch because these falling 
tones are established only secondarily by VIII 4 66. This discussion of Katya- 
yana, in which he points out a technical flaw, suggests a way to remedy it and 
finally decides that Panini himself had already taken care of the problem, is a 
fine specimen of Panimya scholasticism. Under the rule 1 2 37 itself Katyayana 
follows a different course by actually rewriting the whole rule. “In the Siibrah- 
manyd the vowel /o/ is [always] high pitched’’ (varttika 1); “The vowel /a/ is 
[always] high pitched before a verb form; also the first of the following [syl- 
lables]” (varttika 2), etc. This is the technique of the Pratisakhya-s: a mechani- 
cal enumeration of instances instead of a statement of the sahent feature — ^not 
a word on the elimination of the falling tone ( svarita ) ! 

Katyayana’s attitude towards Panini shows great respect; he not only closes 
every last varttika of an adhydya with the auspicious word siddham It is 
correct’ — ^but the whole opus closes with the deferential clause . . . bhagavatah 
Pdnineh siddham “[This formulation] ... of the venerable Panini is correct.” 
The definition of a varttika is in Nagojihhatta’s words “The critique of what 

32 Varttika 3 on I 2 39. 

33 P. Tbceemb, IC 4. 203— 208. A specimen of modem recitation of the Subrahmanya 
is found on the LP record album ‘The four Vedas’ by J. Levy and J.F. Staae 
(Asch Mankind Series, New York, 1969). 

34 Vin 4 66 vdattad anuddttasya svaritah “After a high pitched [vowel], for a low 
pitched [vowel] one with falling tone is substituted” (e.g. brdhmdTmh > hr dhmdimh). 


H. Scharfe * Grammatical Literature 

has not been said or said badly/'^s i. e. an investigation into the correctness of 
Panini’s rules. If found wanting, emendations or additions are proposed; if 
found superfluous, their elimination is recommended. In the majority of cases 
the rules are found to be coirect as they stand ; in fact, Katyayana goes to great 
lengths to save the original formulation of Panini. I find the earliest suggestion 
of a bias against Katyayana in ^abarasvamin’s commentary on Mimamsa-sutra 
X 8 4. ^abarasvamin, who does not accept varttika 2 on Panini II 1 1, rebukes 
the Venerable Katyayana' for not telling the truth (a-sad-vddin) due to igno- 
rance. Th. Goldstiicker went still further in his criticism in his book Tanini: 
his place in Sanskrit literature.' He saw in Katyayana a vicious enemy of 
Panini and in Patanjali a loyal defender. Such views are not tenable. 

The Astadhyayi can be compared to a code of law which is subject to legal 
interpretations^ when cases arise that were not or could not be foreseen by the 
lawmaker. The courts need a consistent and workable application even in such 
cases. Lawyers are used to obtaining this apphcation by extrapolating princi- 
ples embodied in the code which is presumed to be comprehensive and consist- 
ent to the minute technical details ; seemingly redundant features must have 
their significance. If these extrapolations lead to opposing conclusions, this 
contradiction must be resolved. As a last recourse, the law may be amended. 
The Paniniya-s are like such lawyers and we miss the point when we castigate 
them for reading later theories into the original texts.^s But the philologist’s 
objective is different. He wants to know what Panini meant when he formulated 
his rule, comparable to the historian of law who is interested in the original 
intent of the lawmaker and the meaning his work had in his time. Instead of 
being flustered by contradictions that appear in the extrapolations he sees in 
them the sign of a lively scholarly tradition, an indication of problems that 
surfaced only after the author's time. 

The problem of maintaining consistency in such a large and complex struc- 
ture as the Astadhyaju must have been staggering, much more so than it would 
have been in a paradigmatic kind of gram mar. Add to this the irregularities of 
a natural language and it was inevitable that there should be contradictions if 
we attempt to extrapolate the underlying principles. Furthermore, it is in the 

Uddyota I 125, 19 (varttika 1 on I 1 1) sutre ^nuhta-duruhta-cintdharatvam 

35 p . Th x e m e, JAOS 76. 23. It is consistent with this view that interpreting gram- 
marians apply in any difficulty the weakest argument first, escalating the debate 
to stronger and more basic principles only when forced to do so: B. A. vak Noo- 
TEN FoL 7. 598f. with reference to Mbha^ I 286, 1-15. 

37 M. Deshpakde, KZ 86.229, points out how the rules III 4 103 and VI 4 71 
lead in consequence to opposing principles: does I 1 69 apply to a long /a/ or not? 
His paper makes it clear that the Astadhyayi was not constructed with the absolute 
consistency claimed by the interpreters. Whichever way one understands 1 1 69—70, 
there remain some loose ends (see especially p. 245 of said paper). 

33 E.g, S.H. JosHi, Vyakarana Mahabhasya, Avyayibhava-tatpurusahnika 
(Poona, 1969), p. x. 



nature of nnformnlated subconscious principles that they are not quite applied 
with the same consistency as expressed metarules. I might add that our task in 
interpreting Panini is not to further embroider the intricate scholastic patterns 
by new sophisms ( phakMkd-s>^^) but to cut through to the original material: not 
a faultless weave as the Paniniya-s would have it and not quite the automatic 
device they pretend it to be — ^it is but a ‘thread’ and after the loss of the oral 
instruction that once went with it we must know the results (i.e. Sanskrit) to 
check our procedure and to avoid wrong forms.^o 

Seven references in the varttika-s to emendations proposed by ‘some’ (eke) 
show that Katyayana was not the first to investigate the Astadhyayi in this 
fashion.41 Xhe Moka quoted m Mahabhasya II 398, 4 + 13 might well be older 
than Katyayana. To express possession, Panini V 2 115 teaches the suffixes -in 
and -ika for noun stems ending in. I a>l (e.g. davda: dandin, davdika); the 
next rules adds the nouns vrlhi, etc. (vrlJii: vrlhin, vrihika; sikhd: sikhin). 
The verse demands a restriction: sikJiddibhya vrdr vdcya ika'^ yavakhadddisu “-in 
must be taught [exclusively] for sikhd, etc., -ika [exclusively] for yavakhada, 
etc.” Katyayana (varttika 1 on V 2 116) regards the restriction as unnecessary, 
an apparent polemic against the sloka. 

Katyayana attached his varttika-s to Pamni’s sutra-s in their natural se- 
quence in the Astadhyayi; as the authoritative form of the sutra-s was their 
continuous recitation he had considerable leeway in dividing sutra-s into 
two:"^^ e.g. he proposes to read the sutra 1 1 17/18 u^.a um as two: 1 1 17 u^.ah 
and 18 um — an emendation that has been adopted in the vulgate text. In 
other instances Katyayana found a sutra unnecessary. Panini I 1 48 e^,a i^ 
ghrasvddese teaches that /i/ and /u/ are substituted for /e,ai/ and /o,au/, 
respectively, when the substitution of a short vowel is decreed for the latter 
sounds: citra-\-go>citra-gu (I 2 48). The correct substitution can in Katya- 
yana’s opinion however be eiffected without the rule I 1 48 because there is no 
short /e,ai/ or /o,au/ and because /i/ and /u/ belong to the same group and are 
therefore the natural substitutes. In this case Katyayana’s suggestion (in spite 
of Patanjali’s silent consent!) did not prevail in the tradition and the vulgate 
still contains the sutra I 1 48. 

This word is derived by a transposition of the aspiration from Middle Indo- 
Aryan pakkhika (to Sanskrit pdksika ‘siding with one party’) as it typically reasserts 
the refuted purva-paksa [MW ; Adar^ Hindi Ko^ (Benares, 1964)], short for purva- 

Cf. Hagojibhatta, Paribha^endu-sekhara on Paribhasa 9: atra laksyanusdri 
vydkhydnam eva satanam “Concerning this we can have recourse only to the inter- 
pretation guided by the particular forms.” 

F, EIielhorn, IA 16. 103. 

42 Note that there are but a few suggestions (by Patahjali only) that a word at 
the beginning of a sutra may instead belong at the end of the previous rule (Kjel- 
HORN, IA 16.247) which reveals a basic consensus on the division of the sutra-s. 


H. Scharfe • Grammatical Literature 

Proposed changes in the formulation of sutra-s are motivated either by 
logical-technical problems or by additional language material that needed to be 
covered. An example of the logical problems we have seen in the definition of 
sa-mrm (p. 140) ; but they are not limited to definition rules. While we easily 
get vrksau ‘two trees’ from vrhsas ca vrhsas ca ‘a tree and a tree’ (above p. 136), 
Katyayana sees the need for a special injunction regarding synonyms. We 
shoifid be able to combine vahra-danda ‘someone with a bent staff’ and hutila- 
davda ‘someone with a crooked staff’ in a single dual form, either mhra-dandm 
or Jcutila’davd^'^ people with bent/crooked staves’ (varttika 23 on I 2 64). 
Technical simplification is the aim of a varttika on Panini 1 1 34: the listing of 
the seven direction adjectives can be shortened by a reference to the gaim 
‘word list’ attached to 1 1 27.^3 

Regarding the additional language material, we must try to differentiate 
between additional old Vedic forms and new or regional forms. Papini’s presen- 
tation of the Vedic material was highly selective, possibly restricted to forms 
that occurred in commonly used liturgical texts ; for reasons that so far elude us, 
Katyayana included further Vedic material. Panini VI 3 21 allows the retention 
of the genitive suffix for the first member of a compound only when an insult 
was intended (type ddsyah-putra ‘[illegitimate] son-of-a-maid’) — ^Katyayana 
postulates also a non-abusive Divo-ddsa (a name found in the Rgveda) and 
Devdndm-priya (the very recent title of the Maurya kings). Frequently the 
words or forms postulated by Katyayana^ are not attested to in any text which 
makes it virtually impossible to decide between the two kinds of language 
material. Furthermore, it is possible that Panini has overlooked certain words 
or forms or has not thought it necessary to teach them. Forms postulated by 
Katyayana do not, therefore, necessarily indicate linguistic development. 

A major morphological development reflected in the varttika-s is the forma- 
tion of the periphrastic perfect with '^kr, ]/^ and '^hhu, whereas this perfect in 
the language of the Brahmana-s (and Panini) could only be formed with ^kr 
(Panini HI 1 40 with varttika 3): 'pdcaydm cakdrajasalhahhuva. Another ex- 
ample: in agreement with older usage Panini 1111 teaches dvi-vacanarn 

pragrhyam “/i, u, e/ expressing duality are pragrhya” i.e. must be ‘separated’ 
from an eventually following vowel without any sandhi effects. Katyayana 
takes the sutra to mean that dual suffixes ending in /i, u, e/ are pragrhya so 
as to include the verbal dual forms in ~dthe>^ etc. which were not treated as 
'pragrhya in the older language. The aspiration of stops before sibilants {vatsa > 
vathsa, etc.) mentioned in varttika 3 on VIII 4 48 as the doctrine of Pauskara- 
sadi probably reflects a recent popular pronunciation and corresponds to state- 

43 H. ScHABTE, Panioi’s metalanguage (Philadelphia, 1971), p. 49f. 

44 Katyayana introduces new gana-s like Godddi- (varttika 1 on IV 1 175) for 

which Patanjali, Mahabha^ya II 270. 3 f., gives the full list: Kadera, Kerala. 

45 Surprisingly, these verbal forms are not pragrhya in the Katantra (H. Lttdees, 
SPAW 1930.521 =Ph. Ind., p. 701). 



ments in three Pratisakhya-s (above p. 131).46 The word ascarya which Panini VI 
1 147 and Yaska II 24 have only in the value of ‘rare’ has, by Katyayana’s time, 
assumed the meaning ‘wonderful’ (varttika 1 on VI 1 147; cf. classical Sanskrit 
ascarya n. ‘marvel’).-^’ 

Katyayana, who in his Vajasaneyi Pr. used artificial terms extensively, also 
developed Panini’s metalanguage. The elements 4 and 4i used often by Panini 
to denote roots (e. g. ruc-i III 2 136 ; as-ti II 4 52) are taught in varttika 2 on III 
3 108 as and And in the varttika-s 5 to 8 on 1 1 68 he proposes four new 
determinatives^s to mark those nouns in Panini’s rules that denote a) also their 
synonyms, b) only their synonyms, c) only the subspecies or d) both the word 
itself and its subspecies.^^ 

Many of Katyayana’s varttika-s deal with the principles of interpretation or 
construction. Some of these principles have been laid down by Panini himself, 
e.g. 1 4 2 vipratisedhe param kdryam “When there is conflict, the subsequent[ly 
formulated] rule [takes precedence over the operation provided by a previous 
rule]” which is valid in the definition section I 4 1 to 11 2 38: each item in this 
section is allowed only one technical name. Katyayana extends the application 
of this principle through the whole grammar with the exception of the last three 
pada-s (the so-caUed tripadi, cf. above, p. 101), as a convenient mechanical 
device: whenever two conflicting rules^o tend to apply to a form, the rule given 
later in Panini’s grammar prevails. The arbitrariness of this extension is shown 
by the great number of exceptions that caU for the contrary principle when ‘the 
conflict is in favour of the former rule’ (purva-vipratisiddham).^^ The extension is 
in fact quite unnecessary for the proper interpretation of Panini’s rules if one 
admits the validity of several principles of a general logical nature.^^ 

The most important of these is this: a special rule that falls completely within 
the sphere of a general rule supersedes the general rule in its limited area. If 
this were not so, it could not apply at all and its very existence would be point- 
less {apavada overrules utsarga). Another principle refers to the gradual build-up 
of words from a root and a number of suffixes: the operation within a stem 
{anga) takes precedence over an operation caused by a further suffix being 

According to varttika 12 on 13 1, the root list serves the purpose of eliroinating 
faulty, i.e. colloquial, forms like anapayati (v. 1. dnavayati). 

P. Thteme, KZ 78.110. 

The term anubandha ‘tie on’ is first found in the varttika-s, possibly after the 
analogy of the anuhandhya pasu, the animal which is tied to the sacrificial post and 
subsequently slaughtered. 

H. ScHABFE, Panini’s metalanguage, p. 42 f. 

50 In varttika 12 on VT 1 158 Katyayana suggests even the prevalence of a word 
taught later in the same rule when there is a conflict. 

5^ Patanjali, Mahabhasya I 306, 9f., avoids this oddity by reinterpreting the 
word param in Panini’s rule 14 2: not ‘subsequent’ but ‘desired’ (is^) so that the 
rule states “Where there is a conflict that which is desired [takes place].” That 
implies that one knows the correct forms beforehand and is guided by this knowledge 
in the correct application of the grammar. 

52 G. Cabdoi^a, JlPh 1.40-74. 


H. Scharfe • Grammatical Literature 

attached to the stem ( antat'anga overrules hahir-anga, e. g. in the instr. siag. 
fern, paivyd from [patu-{-i]+d instead of "^patuyd),^^ These principles are based 
on sound reasoning and yet their consequent application leads to certain diffi- 
culties which Patanjali meets with additional interpretation rules. The relative 
strength of these basic principles is tested wherever they come into conflict 
with each other: 

“Of (these five kinds of rules, viz.) a preceding (rule), a subsequent (rule), a 
constant (rule), an intra-stem (rule), and a special rule, each following (rule) 
possesses greater force (than any one of, or all, the rules which in this metarule 
are mentioned before it).”S4 

The loss of the original oral explanation attached to Panini’s sutra-s led to 
an intensive search for the principles underlying the grammar so that its rules 
could be properly applied. This search proceeded on the assumption that Panini 
formulated his rules with absolute consistency and parsimony; any apparent 
deviation from these ideals was meant to indicate the validity of another guid- 
ing principle whose usefulness was shown by a number of further applications.^s 
But the assumption of Pardni’s absolute consistency (cf above p. 142f.) and 
parsimony is not correct he states occasionally a general principle even 
though this could be derived from logic or from a textual 'indication/^? as e.g. 
Panini 1 1 56 sthdnivad ddesah . . . “The substitute is like the original . . .’’ The 
recent Paniniya Nagojibhatta (18th c. A.D.) believes, foUowing a suggestion of 
Patanjali, that this seemingly unnecessary assertion is in itself an ‘indication:’ 
that principles established by logic or by textual indications are not universally 
valid (jndpaka-siddkam na sarvatra)^^-~hy which admission of course the basis 
for this whole investigation has been severely weakened. 

53 The same is true for operations within a word vs. operations in a sentence. The 
replacement /i/ > /u/ in the generation of the 3rd. sing, imperative takes precedence 
over /i/>/y/ in external sandhi: pacatli>u]+atra>pacatv aira. That led to a re- 
mterpretation of the terms anfarahga and hahiranga; ‘whose conditions are internal 
and external’ (Nagojibhatta, Paribh^endu-^ekhara on Paribha§a 50 beginning) ; the 
^ contained anga in the only meaning found in the A^tadhyayi 
and the vartti^-s: ‘that to which something is suffixed’ (Panini 1413). Katyayana, 
varttika 5 on III 4 77, applies the antarahga principle to the above-mentioned case 
only as an alternative ; in varttika 4 he proposed to see here a ‘conflict in favour of the 
tormer rule. In the verse quoted by Patanjali (Mahabha§ya I 91, 20) from Gonar- 
VI 3 ’ ^ instead of anfaranga; cf. Nagojibhatta, Uddyota IV 659 (on 

54 Paribha^a 38 in Nagojibhatta’s Paribha^endu-^ekhara; cf. also the verse 
quoted in Haradatta’s Padamanjari on the Ka^ika on I 4 2. 

5 = Cf the statement nailcam prayojanam yog&ramhham prayoiayati “A single 
ajgication does not motivate the introduction of a [definition] rule,” Mahabha^ya 

5 ol j jnapaka for these occurs first in Katyayana’s varttika-s, e. g. varttika 

58 Paribha^a 116 in Nagojibhatta’s collection. 



Of Katyayana’s achievements as a grammarian the addition of new language 
material has already been mentioned (p. l44f.). He also tried to further Pauini’s 
grammatical analysis.®® Irregular formations are given at times in the Asta- 
dhyayi as ready-made words without grammatical build-up, e. g. mdtdmaha and 
'pitdmaha ‘maternal and paternal grandfather’ {IV 2 36) which, besides the 
words mdtr ‘mother’ (nom. sing, mdtd) and pitr ‘father’ (nom. sing, pita), 
evidently contain the adjective maha ‘great.’ Katyayana sees instead in the 
second element a suffix ^dmaha^ (varttika 2 on IV 2 36) — a rather mechanical 
way to account for the admittedly unusual formation. In rule V 3 22, Panini 
lists isolated time adverbs sadyas ‘today,’ parut ‘last year,’ purvedyas ‘yester- 
day,’ etc. Katyayana’s analysis of sadyas as sa- with a ‘suffix’ dya(s) (varttika 1) 
may still be regarded as reasonable in spite of the obvious connection with the 
root ]/d^ and the words for ‘day.’®® But the analysis oipurvedyusmpdrvam.^ a 
suffix edyusu^ (varttika 6) is very artificial: the vowel /e/ certainly belongs to 
the first element of the word (purve-dyus). repJia ‘burr, r-sound’ i s no t ra + 
(suffix) ipha (varttika 4 on III 3 108) but is derived from the root ^riph ‘snarl’ 
attested to in the Veda. These mechanical divisions fall short of the standards of 
Panioi’s functional analysis and remind one of Yaska’s procedure and — carried 
to an extreme — ^the analyses found in the XJnadi-sutra-s. Katyayana makes 
frequent use of the disappearance (lopa ) of words in secondary word formations. 
Already Panini V 3 82 teaches ‘loss of the latter word’ (uttara-padadopa) when 
the latter word of vydgJirdjina ‘tiger skin’ (viz, ajina ‘skin’) is dropped before 
the secondary suffix ka'^ : vydghraka ‘man in tiger skin.’ Katyayana adds the 
t}q)es sdka-pdrthiva ‘vegetarian king’ from sdkadhojipdrtUvah ‘vegetable-eating 
king, ’61 ustra-mukha ‘camel face’ from ustra-mukham iva mukham yasya ‘whose 
face is like the face of a camel. ’6^ While Katyayana saw here the ellipsis of the 
latter member of a lower level compound that existed before the final composi- 
tion, later commentatorssa start with a hypothetical compound Hakabhoji- 
pdrtJiiva and assume the ellipsis of the middle member.®^ 

In his varttika 1 on II 3 19, Katyayana states that the case relation of a noun 
with the verb takes precedence over that with a nominal supplement to the 
verb: though namas ‘homage’ would require the dative of its object {namo 
devebJiyah by Panini 11 3 16), we use the accusative in namaskaroti devdn ‘he 
reverences the gods’-~~the construction of]/kr ‘do’ with accusative prevails over 
the construction of namas with dative. In the varttika-s 1 and 2 on II 3 28 
Katyayana assumes ellipsis of a gerund: in prdsdddt preksate he looks down 
from the terrace,’ he takes ‘terrace’ as the object of a lost gerund in a hypos- 

59 E.g. the term dgama for morphemes that do not alter meaning or function of a 
word or suffix (I 1 20, varttika 5). 

60 W. ScHunzB, Kleine Schriften, 2nd ed. (Gottingen, 1966), p. 806-828. 

61 II 1 69, varttika 8. 

62 n 2 24, varttika 12. 

63 Bhasavrtti, Durghatavrtti, Durgasimha (on Katantra). 

64 H. SCHABFE, KZ 85.219-223. 


H. Schaxfe * Grammatical Literature 

tasized sentence *prdsddam druhya preksate 'after climbing the terrace lie 
looks’ (the example is Patanjali’s). Of grammatical interest is also Katyayana’s 
definition of a sentence as ‘having one verb’ (eka4i^] varttika 10 on II 1 1); in 
nominal clauses the verb ‘is/are’ is supplied (varttika 11 on II 3 1). 

Katyayana^ first formulated for us the basic principles of grammatical 
analysis which allow us to isolate abstract units like roots, suffixes, etc. even 
though these never occur alone in ordinary language. By a two way reasoning 
of anvaya ‘concmrent presence, agreement’ and vyatireka ‘concurrent absence, 
difference’ we can isolate and coordinate word elements and their meanings. 
Patanjali explains what is meant; the word vrksas ‘tree’ suggests a certain 
physical object with roots, branches, fruit and leaves; the word vrksau ‘two 
trees’ suggests likewise roots, branches, etc., but in two specimens. We conclude 
that the constant meaning ‘tree’ is carried by the element vrksa-, the notions 
‘one’ and ‘two’ by the sounds /s/ and /au/, respectively. An attempt to prove by 
the same method that the single sounds carry meaning is abandoned: sounds 
in themselves are meaningless (varttika-s 9-15 on iSiva-sutra 5). In his remarks 
on varttika 15 Patanjali points out that otherwise the similarity of tejoa ‘well,’ 
supa ‘sauce’ and yupa ‘sacrificial post’ would force us either to assume that the 
three words and the objects denoted by them have more common than separate 
properties or that their special meanings are totally expressed by their first 
sounds alone — ^which would leave a meaningless -upa.^^ 

Z 9=1 3 1, varttika 6; of. G. Caudona, ALB 31/32.313-352. 

6® Mahabha^ya I 32.2-10. 

Chapter VII 


While the first purpose of Patanjah’s Mahahhasya is a discusssion of Katya- 
yana’s varttika-s, in the process of these discussions he frequently refers to 
other authors, quoting metrical lines (and a few prose sentences) from their 
works.^ The later commentators uniformly call these fragments sloka-vdrttika-s 
or merely vdrttika-s ; sometimes they ascribe these verses to Katyayana, author 
of the varttika-s, other times they contrast the author of the sloka-varttika-s 
with the author of the varttika-s. The great variety of metres, ranging from 
sloka-s (165 verses) and dryd-s (about 40 verses) to classical metres like varnsastha 
and dodhaka, suggests a number of authors. Kaiyata (11th c. A.D.) reports a 
tradition that the dodhaka stanza found in Mahahhasya 1 484, 17 f. was compos- 
ed by Vyaghrabhuti, and the sloka in Mahahhasya HI 199, 19 is quoted by 
Patanjali as the tradition of the Bharadvajiya-s. The other verses remain 
anonymous for us. 

Some of the verses are devoted to the same kind of critical investigation as 
Katyayana’s varttika-s, and at times Patanjali’s discussion is no more than a 
prose paraphrase of the verses with illustrations added (e. g. on Panini 1 1 19) ; 
the verses are often repeated without interruption after Patanjah’s elabora- 
tions. In other instances they are quoted beside the prose varttika-s of Katya- 
yana; the stanza Mahahhasya II 121, 7f. even refers to Katyayana by name 
and quotes his varttika on Panini III 2 118.2 

Independent critique of the Astadhyaju seems to have been the objective of 
the prose annotations attributed to the Saunaga-s (‘followers of Sunaga’?) who 
are quoted seven times in the Mahahhasya for their remarks ; once their state- 
ment is repeated without attribution, which shows that Patanjah, even where 
he seems to speak for himself, may actually quote earlier authors.® 

Patanjali’s short discussion of Panini VTII 2 58 gives an idea of the rich 
grammatical activity in the centuries between the two authors. The Indo- 
European root "^vid has differentiated in Sanskrit into several homonyms foUow- 

1 F. Kxelhorn, IA 15.228-233. The metrical quotations are conveniently gath- 
ered at the end of vol. 2—5 of the Rohtak edition of the Mahahhasya. ^ 

2 There can be little doubt that bhagavdn Katyah ‘the venerable Katya’ in the 
stanza Mahahhasya II 97, 26f. refers to the same Katyayana. The reference to 
varttika 3 on Panini III 4 37 in the line quoted in Mahahhasya 11 176, 12 contains 
the earliest use of the word varttika for Katyayana’s annotations. 

3 Mahahhasya II 209, 8 (anonymous) =n 105, 8 and 238, 11 (Saunagah). 


H. Scharfe * Grammatical Literature 

ing different inflectional patterns; their part. perf. pass, are vit-ta^ vin-na and 
vid-ita. Patanjali quotes three anonymous stanzas which were attempts to 
arrange this material. The first is a dodhaJca (the first quarter is defective) which 
distinguishes four roots ]/md: after the root ^/vid that follows the 6th verbal 
class the suffix is retained, after of the 7th class /t/ is optionally replaced 
by /n/, of the 4th class is treated like ychid and that of the 2nd class has 
an /i/ inserted between root and suffix. The peculiar feature in this stanza is 
that it gives the character of the 6th class as while Panini calls it that 
seems to indicate that the reference is not to Panini’s grammar but to one very 
close to it. The second stanza is a Moha which repeats the statement in strictly 
Paninian terminology and with greater concision. The third stanza, also a sloka^ 
finally takes the popular approach of pairing finite verb forms with their part, 
perf. pass. : from vetti the p.p.p. is vidita, from vidyate vinna, from vintte both 
vitta and vinna, and from vindati vitta. 

Even Katyayana’s varttika-s had become the object of interpretation and 
critique before Patanjali. The school of the Bharadvajiya-s (‘followers of 
Bharadvaja’) is quoted ten times for contributions on the varttika-s.^ In most 
cases, the varttika has been enlarged to cover additional items, and in a few 
instances there are alterations or even a different solution; once the remark is 
addressed directly to Pamni’s rule. Other remarks on the varttika-s are quoted 
anonymously. Varttika 4 on Panini HI 2 171 postulates the words sdsahi^ 
vdvaJiij cdcali and papati; here Katyayana obviously has been caught napping^ 
because there is no need to list these as ready-made words as their analysis is so 
cl ear: the suffix -i is used after the intensive of the roots i/sah, ]/vah, ycal and 

Occasionally the dispute involves the tradition of the varttika-s themselves. 
In the very involved discussion on Panini 1 1 69 (vowels of the sound table denote 
also the other vowels of their respective classes, i.e. /a/ denotes also /a/, etc.), 
Katyayana says that this is unnecessary because the ‘form’ of the vowel grasps 
the other vowels (varttika 7) and because these are not different from each 
other (varttika 8). With a slight variation in the text of varttika 8, the other 
tradition quoted by Patanjali takes these two arguments as one: “. . . are not 
different because the ‘form’ is grasped” — an easier reading but hardly in 
Katyayana’s style.^ 

In his discussion of why Panini used two determinatives h and n when the 
functions of the two are almost identical, Katyayana supports the existence of 
both with the reference to the root yjdgr which shows the base vowel in some 

^ While most of these quotations are in varttika-like prose, one (Mahabhasya in 
199, 19) is a half-57oA:a. 

s If he had a good reason for his proposal, it has been lost. 

® Mahabh^yan 135, 11. 

Mahabha§ya*I 179, 22. We expect the reason in an ablative, not in an inde- 
pendent nominal clause. 

Fragments preserved in the Mahabha^ya 


forms (e.g. jdgrtah) and guim in others (e .g. jd garitah): jdgro "gwm-vidhih ‘‘[The 
purpose is] establishing non-guTm for ]/jdgr.'' The variant reading quoted by 
patanjali, Mahabhasya I 193, 1, states conversely jdgro guim-vidhih “[The 
purpose is] establishing guTia for ]/jdg^ Factually speaking, both readings 
amount to the same thing. One cannot even properly speak of two readings 
because, with the negation elided, both sound exactly the same. But it is my 
impression that the authors of sutra-s have always avoided formulations where 
controversial negations would be completely obliterated by sandhi; had Katya- 
yana indeed intended the negated form, he would have found a way to tell us 
so more clearly.® 

In another class are the lines quoted in Mahabhasya 11 284, 1® + 14£; 310, 
9-311, 6; 398, 13, which do not discuss Pamni’s formulations but teach gram- The dependence on Panini is obvious^^ but there are also characteristic 
differences. Panuai reduces several taddhita suffixes through the use of hetero- 
phones (TH —ik, CH etc.) which results in metalinguistic forms like THa^, 
TEa^, ^THa^, etc. Our metrical grammar makes no use of these heterophones 
and teaches the suffixes as ika^, ika^, Hka^, etc., otherwise retaining Panini’s 

8 How commentators take advantage of such alleged ambiguities shows m the 
traditional interpretation of Sainkhya-karika 41c tadvad vind [']vi8e8aih* 

® To obtain a correct line, the unnecessary word vidyd must be omitted (F. KrEL- 
HORN, lA 15.233). 

Note the familiar verse fillers smrtah and isyate (twice). 

11 The metrical form makes it very unlikely that this grammar was older than 
Panini because the general tendency was to shift from sutra-like manuals to metrical 

12 The suffixes (atyai > strl) and W ( ?) mentioned in a verse quoted 

in Mahabha 9 ya 1 245, 26 are not Paninian either. On ^avat» (Mahabhasya 11 378, 21), 
ghu (Mahabhasya III 229, 2) and Icb (Mahabhasya H 284, 11) cf. G. Cabdoka, TP AS 
1969. 30 and 35a. 

Chapter VIII 


“Either read the Mahabhasya or rule a large kingdom” (MaMbhdsyam vd 
'pdthaniyarri rmMrdjyam vd pdlaniyam) — ^this saying among the pandits shows 
the high regard in which they hold Patahj all’s work. Modern critics have joined 
them in their praise of Patanjali’s simple yet vigorous style and his sound 
reasoning as well as his vast learning that presents us with the quintessence of 
generations of grammatical-philological research.^ For him, Sanskrit is still the 
spoken language of daily conversation, if only for some classes of 'good society,’ 
and he handles it with evident authority — ^maybe the last grammarian to do so. 
Correct use of Sanskrit could no longer be defined geographically or sociologi- 
cally in his time. That gave Patanjali an extra motive for studying Panini’s 
grammar: as a standard, a device to identify good native speakers of Sanskrit.i^ 
Of the three sages of grammar, viz. Panini, Katyayana and Patanjali, accord- 
ing to a common tradition (e.g. Kaiyata on Panini I 1 29) the later author 
overrules the earlier one in case of a conflict of opinion, yet three Buddhist 
commentators attached greater value to the words of Panini.^ 

In cross references to his own words, Patanjali calls his work simply hhdsya 
'[work composed] in colloquial language’ and indeed it is stylized as a series of 
shorter or longer debates. A student may ask a question which is answered by 
the teacher or his younger assistant; often the answers of the assistant are not 
satisfactory and are set aside by the final opinion (siddhdnta) of the teacher to 
whom both student and assistant turn for help. These labels ‘student,’ ‘teaching 
assistant, ’3 and ‘teacher’ which the later tradition supplies are useful in our 
imderstanding of these debates; they reflect the common school practice as 
observed by D. Ingalls.^ But we must keep in mind that they are superimposed 
on the hhdsya text (which has nothing of this kind), and that often the commen- 

1 SmDESHWAE Varma, VIJ 1.1-36. 

Mbha§. m 174, 6-15; cf. P. Thieme [Felicitation Volume S.K. Belvaxkar], 
p. 60-62. Even now there are native speakers of Sanskrit found in Benares as 
Pt. Kagaeaja Rao (Seattle, Wash.) informs me. 

2 L. Renott, La Durghatavrtti de Saranadeva (Paris, 1940), p. 78, fn. 2: Nyasa, 
Bha§a-\’Ttti, Durghata-vrtti. 

3 dcdrya-desiya ‘one who is almost like a teacher, a not-yet-accomplished teacher,’ 
cf. Panini V 3 67. 

^ In: M. Singer, Traditional India (Philadelphia, 1959), p. 5. 



tators’ very definite opinions as to which is Patanjali’s final view are based 
only on their jndgeirLent of the merit of the arguments and not on a formal 
indication of Pataiijali.s 

Patanjali must have composed his work sometime around 150 B.C. because 
of several references to historical events of his time.® Katyayana, varttika 2 on 
III 2 111, had postulated the use of the imperfect suffixes for something the 
speaker did not witness though he could perhaps have done so and which is 
commonly known. Patanjali, Mbhas. 11 119, 4f., illustrates this amendment 
with two sentences: aruimd Yavanah Sdketam “The Greek besieged Saketa 
( = Ayodhya/Oudh)” and aruimd Yavano Madhyamikdm “The Greek besieged 
Madhyarmka (Chittor).” This is a reference to the raids by the powerful Greek 
kings of Baktria (Afghanistan) , though the exact year of these incidents remains 
unknown. The use of present tense suffixes for a work already begun but not 
completed is illustrated with the sentence iha Pusyamitram ydjaydmah “Here 
we conduct a ritual for Pusyamitra” (Mbhas. II 123, 3f.). This Pusyamitra was 
the founder of the Sunga dynasty that replaced the Maurya-s in 189 B.C. It is 
reported that Pusyamitra twice performed the horse sacrifice ritual (asva- 
medha),'^ and Patanjali’s illustration should refer to one of these performances. 
It is not clear what value should be attached to the reference to §aka-s in 
Northwest India outside Aryavarta (IVlbhas. I 475, 4): when did Saka-s first 
come into India?® 

Patanj all’s home may have been Mathura, which figures prominently in his 
examples, or a place nearby because one travels, he says, to Pataliputra via 
Saketa (Mbhas. II 162, 6f.).® The popular dialect forms which he quotes (Mbhas. 
1 259, 6-14) correspond, with their retention of /s/ and /s/ and the development 
/s/ > /s/, to the dialect of the gambler Mathura in the drama Mrcchakatika 
(Act 11). This deduction is preferable to that of K.V. Abhyankar^® who con- 
cluded from astronomical data contained in the text that Patanjali lived north 
of Taxila and west of Shrinagar. Not being an astronomer himself, Patanjali 
would have taken this information from other works, and his praise of the 
speech and the customs of the people in Aryavarta would be inconsistent with 
his residence outside this hallowed province. 


5 B.G. BHAnuAHKAB, lA 5.345-350; F. Ejblhobn, Katyayana and Patanjali, 
p. 52 fh. and lA 15.80f. 

6 D.C. SmcAB,, IHQ 15.633-638, discounts these references and proposes the 
2nd c. A.D. as a likely date for the Mahabha^ya. 

Epigraphia Indica 20. 55-57 (Ayodhya inscription of Dhana[deva]) and Kali- 
dasa’s Malavikagnimitra, Act V. 

8 E. FRATTWAinNER, WZKSO 4. 108-111 and S. Chattopadhyaya, The Sakas in 
India, 2nd ed. (Santiniketan, 1967), p. 11-20. 

9 H. ScHARPE, JAOS 96.274. 

10 Select Critical Notes in his 3rd edition of Kielhorn’s Mahabha§ya text, vol. I 
(Poona, 1962), v. 571 f. 


H. Scharfe • Grammatical Literature 

Later traditions (e.g. Eamabhadra Diksita’s Patanjali-caritaii) identify this 
Patahjalii 2 ^th the supposed author of the Yoga-sutra and also ascribe to him 
a lost work on medicine. But the same name, admittedly rare, is not sufficient 
reason for this identification; we even hear of two more Patanjali-s in Sanskrit 
literature. The language of both works is quite differentia and the philosophical 
background of the Yoga-sutra (a compilation which has probably more than 
one author anjrway) shows no more similarities with that of the Mahabhasya 
than can be expected in texts of this period. Modern scholars are divided on this 
question.14 Tradition regards Patanjali as an incarnation of the mythical snake 
Sesa and depicts him as a snake from the waist down.i^ 

The concluding verses of the second book of Bhartrhari’s Vakyapadiya give 
us some idea of the vicissitudes the tradition of the Mahabhasya had to undergo. 
After a period of sophistry, the oral tradition in South India was interrupted 
until it was revived by the Buddhist scholar Candragomin (5th c. A.D. ?) who 
relied on the North Indian tradition (probably of Citrakuta/Ramagiri).i® The 
text itself has come down to us in excellent condition. Its principal division into 
8 chapters with 4 p^-s each follows Panini’s grammar, whereas the division 
into 85 daily lessons’ (dhniha) of about equal length (never crossing pdda 
lines) is nowhere mentioned in the text itself and may be a later addition. 
According to the Chinese pilgrim I-tsing, advanced scholars learned the 
Mahabhasya in three years.!"^ 

The Mahabhasya is, in the first instance, a commentary on Katyayana’s 
varttika-s. But Patanjali did not stop at this and investigated Pamni’s formu- 
lations on his own. Altogether his investigation covers 1713 sutra-s of Panini 
not counting those mentioned incidentally in the course of these investigations. 
A complete explanation consists of a separation of the words, example, counter- 
example and supplementation of the sentence (Mbhas. 1 11, 22 f.). But such an 
explanation of the varttika-s (e. g. Mbhas. 1 30, 2-6) is often only the beginning 
of Patanjah’s own investigation. This may take the form of a simple addition 
as in his comment on Panini III 3 130 (according to Panini’s rule, the suffix 

Edited by Pandit Sivadatta and EIasinath Pandurang Pabab (Kavya-mala 
51 ; Bombay, 1895), verse V 25. 

The assumption is wrong that Gonika-putra and Gonardiya, two authors 
quoted in the Mahabhasya, are Patanjali himself [E. Kiedhobn, IA 15. 81-84 and 
P.S. SuBBAHMANYA Sastbi, Lectures on Patanj all’s Mahabhasya, I, 2nd ed. 
(Thiruvaiyaru, 1960), p. xl-xliii]. Not refuted by S. Li6vi [Fs. As. Mookebjee] 
11. 198 =M5morial S. Li^vi, p. 307. 

13 L. Renou, IHQ 16, 586-591. 

I'l M. Eliade, Yoga (New York, 1958), p. 370f.; B.N. Pubi, India in the time of 
Patanjali, 2nd ed. (Bombay, 1968), p. 15-18; B.K. Matilad, Epistemology, logic, 
and grammar in Indian philosophical analysis (The Hague, 1971), p. 104, fn. 10. 

15 B.N. Pubi, India, p. 2, fn. 3; J.F. Staad, A reader on the Sanskrit grammarians 
(Cambridge, Mass., 1972), p. xvif. 

15 H. ScKABEE, JAOS 96.276. 

17 J.F. Staax, a reader, p. 15, 



.ana found after certain roots with isad, dus or su ‘‘is found also after others”). 
A varttika specifies: the suffix -ana is added to the roots Y^ds, ^yudh, ^d^ and 
ydim, in colloquial language only. Patanjali spells out the resulting words: 
Duhidsana, Duryodhana, dwrdariana, durdharsana. Then he adds: “‘Also after 
the rooty mrf must be added [to this list to account for the word]: durmarsaTia'' 
(Mbhas. II 157, 9-12). 

The following discussion goes much deeper. Panini III 1 67 teaches the suffix 
•ya to denote the object or the existence of the action (personal and impersonal 
passive in our terminology). Patanjali asks why the dual or plural forms may 
only be used when the verb denotes objects (e.g. facyante odandh ‘rice paps are 
being cooked’) but not when it denotes the existence of the action [dsyale 
hhavatdibkavadbhydm ‘sitting is being done by thou/you two’). The answer is 
that the objects may be many but the action is only one; if action nouns like 
'pdha ‘cooking’ appear occasionally in the dual or plural {'pdhau ‘two acts of 
cooking’) this is due to different material substrata. Thus the problem remains, 
viz., that the same logic should apply to actions expressed by passive verbs, 
and Patanjali tries another approach. When we consider different acts of 
cooking, he says, e. g. the cooking of rice pap, the cooking of molasses and the 
cooking of sesame, the three words ‘cooking’ are the same and can be summed 
up in a plural noun form the same occurs (but only rarely) with verb forms 
(e.g. hatasdyikdh iayyante “The slain men lie in different postures,” lit. ‘lyings 
of the slain take place’). If we however consider time differences, the nouns (e.g. 
‘yesterday’s cooking,’ ‘today’s cooking’) each are stfil identical and can be 
contracted into a single dual or plural form, while the verb forma (Hes, lay, will 
lie) are different and hence not eligible for comprehensive plural forms. Patan- 
jali concludes this argument by listing the differences between an action ex- 
pressed by a nominal suffix and that expressed by a verbal suffix. The former 
becomes like a thing which means that another action can inhere in it or 
(grammatically speaking) that the noun can be construed with a verb (‘cooking 
happens’); it also can be a factor in another action (‘rice is prepared by cook- 
ing’). The opposite is true for an action expressed by a verbal sxiffix: no action 
can inhere in another action. Furthermore, an action expressed by a verb shows 
time, person and voice, and it calls for an expression of the agent; not so an 
action expressed by a nominal suffix. The use of gender and number is at least 
partially different. 

Now Patanjah turns his attention to Katyayana’s first varttika. Shall we 
assume that in a verb existence, object and agent of an action are expressed 
by the personal endings or by the stem-forming suffixes Panini’s rules could 
be taken either way. If the personal endings are taught to express existence, 

Following Panini I 2 64 sa-Tupdndm elca-sBsa eka-vibJuikUiu “Only one of the 
identical [words] remains if followed by the same case suffix:” pdkas ca pdkas ca > 
pdkau, etc. 

E.g. in hhav-a-ti ‘becomes’ -a- is the stem-forming suffix, -ti the personal 


H. ScKarfe * Grammatical Literature 

object, etc. it is feared that the rule may no more be a restriction on the use of 
singular, dual and plural forms (varttika 1). But if the stem-forming suffixes 
should express these notions it is feared that these suffixes may in some cases 
not even come into existence because their role has already been pre-empted by 
a nominal suffix (e.g. dharaya folding’ formed from a causative stem with a 
suffix -a denoting the agent; varttika 2).2o Besides, several verb forms have no 
stem-forming suffix at all (e. g. in the perfect tense) or lose it in the grammatical 
process (e.g. the verbs of the 2nd class). Patahjali concludes that it is the 
personal endings that denote existence/object/agent of the action and he settles 
the objection raised in the first varttika with a quoted stanza: just as the 
nominal case suffixes denote both action references (such as object, instrument, 
etc.) and number, a single rule can teach the personal endings to denote 
existence/object/agent of the action and at the same time impose a restriction 
on the use of singular, dual and plural forms.^i 

With great stylistic art Patanjali has created the impression of a freely 
progressing debate with new disputants butting in now and then in which aU 
possibilities of an interpretation are scrutinized. Clearly not every argument 
made (a few were omitted in the preceding summary) carries Patanjali’s con- 
viction; even the end of a debate may not necessarily indicate his preference .22 
The metarule nirdisyarndTiasyddesd hhavardi^^ says that substitutes can only 
be applied to forms that are actually enounced in a rule, and Patanjali makes 
frequent use of it (e.g. Mbhas. 1 29, 23; 116, 19; III 50, 18). And yet he has re- 
futed it in two detailed discussions (on Panini I 1 49 and VI 4 130) because it 
would render Panioi's own given metarule 1 1 49 sasthi sthane-yogd '‘The geni- 
tive is to be construed with a supplied 'in place of . . redundant. 

Patanjali’s statements reflect a further development of the Sanskrit language. 
To give but one example, singular dvandva compounds which were formerly 
limited to types hke 'pdni-'pddam 'hand and foot’ can now be formed from any 
kind of constituents (Mbhas. I 232, 4f.).24 The exemplary speech area is no 
longer the North but Aryavarta, the 'hub of the Arya-s,’ i. e. the central portion 
of the North Indian plain. The Arya-s are contrasted with the residents of 
Afghanistan (Kamboja; cf. Nirukta II 2) and such Indian provinces as Surastra 
and the pracya-madhya 'eastern central’ (Mbhas. I 9, 25-27). The colloquial 
Eastern forms with /r/ > /!/ (he ^layah for correct he ^ ray ah 'hey strangers!’) are 

20 Being on a lower hierarchical level, the stem-forming suffixes cannot compete 
with Icrt suffixes. 

21 Mibh^. n 58.16-23. 

22 Cf. B. Geigee, SAWW 160 (1909), fasc. 8, p. 6f. and P. Thieme JAOS 76.12. 

23 The 12th metarule in Nagojibhatta’s collection ; first attested to by Katyayana 
in v^tika 3 on VI 4 130 (cf. varttika 4 on VT 1 13). 

24 Other new forms may only be school jokes: puputriyisati, putitrlyisati, putrl- 
yiyisaii (HI 8, 2 If.), all meaning 'he desires to wish a son;’ Candravitti V 1 8 
putrlyisisati, Kasika VI 1 3 puputitrlyiyisati and Mugdhabodha XXI 18 puputitrl- 
yiyisisati carry it even further. 



‘barbarian’ {mleccha; Mbhas. I 2, 8). Even in the exemplary area, good 8peecb25 
is not tiniversal; a Brahmin’s wife or daughter may use colloquial popular 
forms [Ltaha for correct Rtaka, Mbhas. 1 19, 21f ). Only the virtuous educated 
Brahmins of Aryavarta are authoritative native speakers of Sanskrit and the 
study of Panini’s grammar makes it possible to identify such men (Mbhas. Ill 
174, 6-15). The superiority of the educated native speaker over a mechanical 
studied grammarian is illustrated by an anecdote told by Patahjah, Mbhas. 

“For thus spoke some grammarian: ‘Who is the urger-on (pravetr) of this chariot?’ 
The charioteer-cum-bard (suta)^^ said: ‘Sir, I am the driver (prdjitr).'* The 
grammarian said: ‘Wrong wordl’ The charioteer said: ‘Your Excellency^? knows 
[only] what should result [from a mechanical application of the rules of grammar] 
but not what is desired [by good speech usage]: such-and-such a form is desired.’ 
The grammarian said: ‘Hey, are we obstructed by this ill-woven (dur-ula) one?’ 
The charioteer said: "suta is not formed from the root vc" {‘weave:’ m >suta 
‘well woven’) but fromtheroot5w(‘ drive’). [Therefore:] If one wants to scold the 
driving one must say: duhsuta ‘bad suta\ ” 

With Patahjali the interpretation of Panini’s rules reaches a new level of 
sophistication.28 Often he can maintain Panini’s formulation where Katyayana 
had to resort to alterations, and much more than Katyayana he has us rely on 
our knowledge of the Sanskrit forms when we must choose between alternative 
interpretations of grammatical rules.29 His basic assumption is that “it is 
impossible that [in Panini’s work] even a [single] soimd should be without 
[specific] purpose” ; for Panini (no ordinary teacher but an authority) took great 
pains with his composition (Mbhas. I 39, 10-12; cf. the quoted verse III 64, 4). 
Though Patahjali never calls Panini a rsi ‘seer’ as the later tradition does, even 
according to him Panini ‘saw’ the grammatical relations (e.g. Mbhas. I 43, 9). 
The same is said about Katyayana when Patahjali explains why this author 
composed two separate varttika-s 9 and 10 on Panini VHI 2 6 when he could 
have achieved his purpose with a single varttika: 

“First the teacher saw this: . . . [varttika 9] . . . ; that was recited. At a later time 
this was seen: . . . [varttika 10] . . .; that was also recited. And teachers do not 
in such cases retract rules after they have given them” (Mbha?. HI 393, 1-3). 

25 Note that Patahjali does not yet use sarnshrta ‘properly prepared, fit for ritual 
use’ to denote this speech. The word is first used so in the Ramayana V 28, 18 
^ '(vdcam . . . aamshrtam). 

25 A king’s charioteer was often his friend and bard (cf. Krsna and Arjuna). The 
Kautaliya Artha^astra III 7 29 classifies the suta as a special class of brahmin. 

2? While the first address (dyusmant) was appropriate for a charioteer toward 
the rider of the chariot (cf. Bharatiya Natya-^astra XIX 11), the second (dei?dna?n- 
priya, the title of the Maury an kings) is ironical (H. Schabfe, ElZ 85. 212). \ayu^rnant 
also Mbhar. Ill 19, 9.] 

25 This was not always appreciated by outsiders who even called grammar a 
‘fatal disease’ (K.C. Chattbbji in: J.F. Staal, A reader, p. 297). 

29 F. Kielhobn, IA 16.246-247; L. Renoij, La Burghatavrtti, Introduction, 
p. 129-135. 


H. Scharfe • Grammatical Literature 

The authoritative teacher can no more retract his ‘true vision’ than Durvasas 
his curse. From this authoritative character of Panini’s and Katyayana’s work 
follows the first rule of mterpretation (and indirectly all others): 

'‘The specific sense [of an expression in a sdstra] is obtained from interpretation ; 
for there cannot be [assumed to be given a definition which is a] non-definition 
[just] because there is a doubt [as to its specific sense]” (Mbha?. I 35, 16 f.). 

Patanjali’s mterpretation rules are either appKcations of a generally accepted 
principle or they are technical explications indicated and necessitated by the 
wording of Panmi’s rules. But even general principles and Panini’s own indica- 
tions are subject to critical discussion. It seems that according to Katyayana a 
special rule supersedes a general rule only when there is no possibility of both 
taking effect together, but according to Patanjali a special rule sets aside the 
general rule in all cases by the popular maxim: 

“When it is said: ‘Let curds be given to Brahmins, buttermilk to [the Brahmin] 
Kaundinya’ ; Kaundinya is given only buttermilk although it would be possible 
to give him both” (Mbhas. 1 115, 1— 4).3o 

Attempts to see in the wording of Panini’s rules ‘indications’ for further inter- 
pretation rules were even more open to challenge. Many are only forced manoeu- 
vres to maintain Panini’s formulation in the face of proposed changes. The 
metarule that intra-stem (antar-aiiga) operations prevail over such as are 
dictated by outside (hahir-anga) forces is both HnguisticaUy sound^i and indi- 
cated by Panini’s procedure. But a difficulty arises in the formation of the gerund 
where the root substitution dhd>]ii (in Tiitvd ‘having put’) is voided when the 
suffix -tvd is replaced by -ya due to an outside force (viz. the prefix) : fra-dhaya. 
Panini’s silence on this point could be an oversight (which is unacceptable to 
Patanjali) or it could imply that in his opinion the antaranga metarule is set 
aside when a gerund substitution tvd >y(i is concerned. Patanjali sees proof for 
the correctness of the latter assumption in the formulation of Panini II 4 36 
where the root substitution od > is taught before both gerund suffixes 

(hence jagdhvd and prajagdhya) : if the gerund substitution would not normally 
set aside the (iTitdTdugci root substitution it would have been unnecessary to 
expressly mention -ya here, but if it does its mentioning becomes necessary. 

Discussions of this kind constitute a major portion of the Mahabhasya. A 
sound principle of interpretation, the presumption that the author’s words 
should make sense, is vitiated often by a belief in the near infaUibility of this 
author (if we take an historic view) or a preference for ad hoc created metaru- 
les^^ over amendments to the basic text (if we take the more appropriate syste- 

30 H. ScH^FE, Die Logik im Mahabhasya (Berlin, 1961), p. 53f.; K.M.K. 
Sarma, Panini Katyayana and Patanjali (Delhi, 1968), p. 170f. 

31 Cf. the modem concept of ‘immediate constituent analysis’ (R. Wells, Lu. 23. 

81-117, and S.D. Josm, JTJP 27. 165-173). ^ 

32 Both roots meaning ‘eat’ supplement each other in the verbal paradigm. 

A recent study on such metarules is A. Wezler, Paribhasa IV, V und XV 
(Bad Homburg, 1969). 



matic view). Patanjali’s practice maugnrated a dubious hunt for further meta- 
rules and their indicators in the text of Panini's grammar until Nagojibhatta 
pruned this exuberant growth back to the level of the Mahabhasya. Patanjali’s 
conservative interpretation has not prevented him from designing numerous 
alterations and the proposed eliramation of whole rules and “in many cases his 
criticism is more thorough-going and destructive than Katyayana’s and . . . 
Panini has suffered more at Ms hands than at those of the Varttikakara.”^ 
Several of Panini’s own metarules are declared redundant (e.g. Panini 1 1 56 
“A substitute is like the original . . Mbhas. I 133, 17-134, 9) because they 
are nothing but applications of universal principles. But twice (Mbhas. 1 14, 7 
and 39, 8) PatanjaK declines to revamp the whole system of metalinguistic 
markers (determinatives) with the words: sidhyaty evam a-Pdninlyam tu bhavati 
“It works this way but it becomes un-Paninian.’’ 

Another prominent portion of the discussions deals with the dittoing process 
inherent in the sutra-style. When an expression is dittoed through a string of 
consecutive rules, a problem arises in case its absence in one of the intervening 
rules is essential. All remedies suffer from the arbitrariness involved: the dit- 
toed expression may leapfrog’ or the whole rule in which it is enunciated may 
be dittoed to serve as a neutralizing vehicle. I shall illustrate both procedures 
with the proposed emendation of Panini 113 i^.o gwm-vrddU which Patanjali 
proposes to reduce to i^.dh (Mbhas. 1 44, 3-8) : 

111 vrddMr ai^ mddhir d^ ai^ 

2 e" gunah [vrddMr leapfrogs) e" guTmli [vrddMr d^ ai^’l 

3 i^.ah \guna}}\ [vrddMrl i^.ah [gunah] [vrddMr] 

In his discussions Patanjah shows an unusual resourcefulness^^ and displays 
his familiarity with the methods and doctrines of the Mimamsa, Samkhya and 
early Nyaya-Vaisesika.^® There are also frequent references to law and custom. 
Illustrations and sample sentences are carefully chosen and represent charac- 
teristic situations. He demonstrates the transformation of cases in a sequence 
of sentences: 

34 F. Ktelhobn, Katyayana and Patanjali, p. 52. 

35 Sometimes a less sympathetic reviewer is tempted to speak of ‘tricks,’ e.g. 
when Patanjali proposes to interpret Panini 115 Jc-n-it.i ca as *JC’-k-n-itd ca standing 
for *g-]c~n-it.i ca to meet a technical difficulty (Mbhas. I 269, 10—12). Certainly that 
was not Panini’s intention; a more serious objection would be that the interpreta- 
tion depends on information not contained in the Astadhyayi. 

36 One should not ascribe to Patanjali the authorship of all or most of such 
doctrines; cf. P.V. KAnie, History of Dharmasastra, vol. V (Poona, 1962), p. 
1156—1158. The two types of negation (nominally bound and verbally bound) have 
recently attracted much attention: L. Ben’OTJ, La Durghatavrtti, introduction, 
p. 114f.; H. SCHABFE, Die Logik, p. 63f.; J.F. Staae, BSOAS 25, 58-61 and JAOS 
83.255; G. Cabdon-a, Lg. 43.34r-56; B.K.Matilal, The navya-nyaya doctrine of 
negation (Cambridge, Mass., 1968), p. 156f. and Epistemology, logic, p. 163. For the 
peculiar reasoning ‘the aorist marker replaces ^Li if it does not replace ^Li see 
H. ScHAarE, Die Logik, p. 85f. 


H. Scharfe * Grammatical Literature 

‘‘The cattle, the horses, the gold are Devadatta’s (genitive) ; the son of a widow 
(nominative) is rich. It is understood: Devadatta (nominative)” (Mbha?. I 264, 

The use of the rare word vaidhaveya 'son of a widow’ is pointless if it should 
refer to any man w^ho has lost his father, or an illegitimate child of his widowed 
mother (or the offspring by niyoga ‘levirate’) as it is hard to see w^hy he should 
be typically rich. But if Patanjali refers to a man whom a widow has adopted, 
everything falls into place: "Indeed, especially in Western India it is notorious 
that when vddows adopt, their most common motive is to take property out 
of the hands of their late husbands’ relations.’’^? We learn from Patanjali’s words 
that this custom was known already 2000 years ago. From similar casual refe- 
rences we can infer (to mention only two more items of interest's) thatPatanjah 
knew the chess gamers and silk reeling.^o Eegarding literature, many references 
point to the existence of hdvya-B^tjlo poetry and the shadow play;4i Vedic 
quotations abound.'^^ 

Patahjali’s grammatical terminology shows a marked return to Paninian 
terms compared to Katyayana’s more frequent use of non-Paninian terms.^s 
In his explanation of varttika 15 on Sivasutra 5, Patanjali replaces Katyayana’s 
afdya with lofa, his upajana with dgama and his vihdra with ddesa (Mbhas. 
I 31, 14-18; cf. 1 202, 21 f.). To account for the word hJilruka ‘afraid’ Patanjali 
teaches a new suffix with determinatives: ^ruka^ (Mbhas. II 135, 14), in analogy 
to Huka^ in Panini III 2 174. 

Patanjali’s contributions as a grammarian are both technical and philosophi- 
cal. An example of his technical contribution is the discussion of the object in 
the sentence grdmam gantum icchati “He wishes to go to the village” (Mbhas. II 
15, 8-11). If both 'going* and ‘village’ are objects of ‘wishing,’ we cannot apply 
Panini II 3 12 which allows for objects of ‘going’ alternatively the dative: 
grdmdya gantum icchati. If ‘village’ is the object of ‘going’ and ‘going’ in turn 
the object of ‘wishing’ we cannot properly construe the passive sentence isyate 
grdmo gantum "The village is the object of a desire to go there.” The solution is 
that ‘village’ is the object of ‘going’ and both ‘village’ and ‘going’ are the 
objects of ‘wishing.’ 

37 J.D.M. Derbett, Hindu law, past and present (Calcutta, 1957), p. 149. 

38 From the many illustrations with Devadatta, the Indian John Doe, StnxxjMAR 
Sen, IL 12. 189-196, has drawn a social portrait. 

3® P. Thteme, Indological studies [Fs. W.N. Brownt] (Hew Haven, 1962), p. 

40 H. Scharfe, Untersuchungen zur Staatsrechtslehre des Kautalya (Wiesbaden, 
1968), p. 330-332. 

41 H. LtiDERS, SPAW 1916.698-737 =Ph. Ind., p. 391-428. 

42 J. Wackernagel and A. Dbbrttnner, ICZ 67.178-182; L. Renou, JA 241. 

43 We have to keep in mind that Patanjali incorporated in his work material from 
many sources which may explain some terminological inconsistencies. 



When somebody, Patanjali says, told a weaver: “Weave a cloth from this 
yam!'' the weaver demurred facetiously: “If it is a cloth it need not he woven, 
and if it must be woven it is no cloth— the words ‘cloth’ and ‘to be woven’ are 
contradictory.” Patanjali’s solution to this problem is that the speaker intended 
a ‘potential name’ (bhdvini samjna), i. e. a name that will be realized at a later 
tune: weave that which when woven will be called ‘cloth!’ (Mbhas. 1 112, 10-13). 
Patanjali here wrongly limits the use of the word ‘cloth’ to a representation of 
an existing object and neglects its representation of a mental image.^ Unsatis- 
factory also is Patanjali’s explanation for the use of the plural ‘we’ by a single 
speaker who sees him self as representing a class (Panini 1 2 59) : “Sometimes one 
wants to express these sense organs as independent entities ... In that case 
there will be [a] plural [form]” (Mbhas. 1 230, 23-231, 2).45 
Already Katyayana had stressed that the study of grammar is a sacred 
obligation; even correct speech usage is meritorious only if it is also theoreti- 
cally understood. Patanjali supplies, apparently from traditional sources, two 
lists with special motivations f6r the study of gramm ar. The first (Mbhas. 1 1 14) 
is short and abstract: protection of the Veda-s, adaption of formulas, traditio- 
nal duty, convenient grasp of facts and removal of doubts. The second list 
(Mbhas. 1 2, 3-6) contains 13 points built almost exclusively on quotations from 
Vedic texts and referred to by the first words of each quotation. 

The benefits of such grammar study are said to be substantial. Patanjali 
insists that the list of sound abstracts at the beginning 

“must be recognized as a mass of hrahmariy full of blossoms and fruit and adorned 
as long as there are the moon and the stars; and in their knowledge lies the 
attainment of the fruit of the merit [obtainable by the study] of all Veda-s, and 
his parents thrive in the heavenly world” (Mbha§. I 36, 16-18). 

The ‘blossoms’ and ‘fruits’ (explained by Bhartrhari in his Mahabhasya-dipika, 
p. 92, as visible prosperity and invisible beatitude) go ail the way back to 
Bgveda X 71, 5 where, for certain people, Speech is said to be lacking in blossom 
and fruit (cf. also Nirukta I 20). 

On these mental images cf. Vakyapadiya HI 3 39; 7, 2-7 and 105 (below, p. 
172 f.). 

p, Thteme, KZ 79. 10. 



The need for a new grammar first arose among the Buddhists after some 
sects had adopted more prestigious Sanskrit versions of their canonical texts. 
While the Buddhists were familiar with most of the nouns in their vernacular 
garb, they lacked the Brahmins’ command of the Sanskrit morphology and 
morphophonemics. They required a simple practical grammar for these topics; 
scientific interest in Sanskrit developed only later. 

The first practical grammar we know of was the Kaumaralata (so named 
after its author the litterateur Kumaralata) of which manuscript fragments 
dating from about A.D. 325 have been found m Turkestan.^ Its terminology 
strives for brevity and shows the influence of writing {bindu ‘drop’ for the anu- 
svdra, bindu ‘two drops’ for the visarga, reflecting the shape of the letters). Just 
as Panini has special rules for Vedic forms, Kumaralata makes allowances for 
peculiar forms of the Buddhist scriptures that resulted from their transposition 
into Sanskrit from Middle Indo-Aryan dialects (e.g, bhdveti for bhdvayati, 
bhesyati for bhavisyati and ehsions of final -aml-im). The name used for these 
forms, drsa ‘belonging to the f5^-s, archaic,’ suggests a reasonable timespan 
between the establishment of the Sanskrit canon and the composition of the 

Probably a recast of the Kaumaralata is Sarvavarman’s Katantra ‘Small 
Manual’ (oldest manuscript fragments from Turkestan c. A.D. 400), ^ also 
called Kaumara or Kalapa.s It consisted originally of four books with four 
chapters each;^ but later the simple grammar was developed into a full-fledged 
system — ^the third and fourth books were merged and another author (variously 
identified as Katyayana, V araruci or Sakatayana) added a new fourth b ook (on pri- 
mary noun suffixes). More additions, some of themin57o^aform, were inserted into 

1 H. Luders, SPAW 1930.502-532 =Ph. Ind., p. 681-714. 

2 H. LtiDEBS, SPAW 1930.482-538 =Ph. Ind., p. 659—721. Even these fragments 
already show insertions. 

3 According to the Tibetan tradition of Taranatha, the author was a South Indian 
Brahmin called Saptavarman. 

4 The original text has been constituted and translated by B. Liebich, Zur 
Einfuhrung in die incflsche einheunische Sprachwissenschaft, vol. I (Heidelberg, 

The Buddhist Sanskrit grammarians 


the other books to cover secondary noun suffixes, composition and feminine suffi- 
xes. In this enlarged form (which by A. D. 800 had found its way into the Tibetan 
Tanjur) the Katantra was the model for the topical arrangement of many later 
grammars. Much later additions are a Dhatu-patha (modelled by Durgasimha on 
that of Candragomin; another spurious Dhatu-patha is found in Tibetan trans- 
lation), ^ a Linganu^asana and an Unadipatha, not to mention a large scholastic 
literature. In its terminology, the Katantra often returns to Panitiian expres- 
sions but makes much less use of metalinguistic determinatives (most of them 
not even expressly defined) and there are no contractions. It lacks the genera- 
tive tendency of Panini’s rules and appears more like a contrastive tabulation. 
The 180 verbal endings which Panini derives from 18 base forms through sub- 
stitution rules are here given in full (III 1 24-33). Sarvavarman was followed 
in this by the Pali grammarians, and by Vopadeva and Anubhutisvarupacarya. 
Type-names like agni 'stems ending in sraddhd 'feminine nouns ending in 
-a,’ dd '^dd and ^dJid' come close to a paradigmatic structure.® The ablaut 
scheme is simpler if less correct: guna arjejo, vrddhi drlai/au (IV 4 34f.).7 The 
formulation of several rules shows the author’s knowledge of the Mahabhasya.® 
The infiuence of the Pratisakhya-s is visible not only in the phonetic terms but 
also in the frequent notion of morphophonemic 'change’ (expressed by the 
accusative) ; the lack of a fixed word order is curious: besides the dominant type 
td nd 'Hd becomes nd'' (II 1 53) there is the type an sas ''sas becomes an'' 
(n 3 9).’ 

An old commentary probably by Sarvavarman himself (of which we stiU 
have reflexes in the fragments from Turkestan) was used by the Buddhist 
author Durgasimha (between the sixth and eighth cent.) for his vrtti; this vrtti 
contains also varttika-hke annotations which try to expand the grasp of the 
sutra-s by addition or interpretation. Durgasimha is also the author of a sub- 
commentary (tlhd). The once very popular Katantra has left its traces on many 
later grammatical systems (Kaccayana’s Pali grammar, Hemacandra, et al.) 
but retained its popularity only in Kashmir, Nepal and parts of Bengal, with 
a large volume of secondary literature. 

5 P. Thibme, OLZ 35.239f.; G.B. Palstile, The Sanskrit Dhatupathas (Poona, 
1961), p. 36-41 ; 49-53. Not much different is the Kasakrfcsna Dhatu-patha recently 
discovered together with a Canarese commentary ; in fact the whole Ka^akrtsna 
grammar was just a version of the Katantra (Palsttub, Sanskrit Dhatupathas, 
p. 44-49 and 17th AIOC, 1953, p. 349-355). 

® The source of this development is Panini I 4 3 i/w stry-dkhyau nodi Feminine 
nouns ending in iZ-iZ are called rwid'i.” 

^ Cf. above, p. Ill, fn. 102. Of little value are simplified Mrafea defimtionshke 
n 4 14 yah karoti aa kartd “Who acts is an agent.” 

8 P. Thieme, OLZ 36.239. 


H. Scharfe • Grammatical Literature 

In a different class is the Candra-vyakarana of the Buddhist litterature 
Candragomin who lived probably c. A. D. 450.^ The influence of theMahabhasya 
is evident in his work at every step. Candragoudn’s promise in his opening 
stanza to produce an easy, clear and complete grammar seems to contrast it 
with Panini’s complicated and Sarvavarman’s eclectic creations; the format of 
eight books with four chapters each is meant to rival Panini’s work. The seventh 
book (on Vedic rules?) and the eighth (on accentsio), however, are missing in 
our manuscripts even though an occasional rule survives as a (quotation in the 
author’s own vrtti.i^ Evidently the Buddhists were less interested in these 
rules that had no application to Sanskrit as the Buddhists used it.^^ 

Besides the sutra-s (c. 3100 in the six surviving books) the grammar consists 
of a Vrtti, a Dhatu-patha with ten root classes, Unadi-sutra-s (arranged alpha- 
betically by the last sound), siksa-like Varna-sutra*s (phonetic statements) and 
86 Paribhasa-sutra-s (metarules). The grammar was widely believed lost and 
its recovery towards the end of the 19th century is owed largely to the labours 
of B. Liebich. The manuscripts (almost exclusively from Nepal, a fragment from 
Kashmir) have remained precariously few for the constitution of a reliable 
text;i4 Tibetan translations have been of some help. The main problem is the 
attribution of the vrtti to Candragomin himself, which seems indicated by the 
lack of any separate introduction or colophon (which merely refer to book and 
chapter of the Candra-vyakarana), and by the use of the first person in nine 
cross-references to sutra-s. On the other hand, K. C. Chatterji (in the notes to 
his edition) has drawn attention to five cross-references in the third person^^ 
and other small inconsistencies: the vrtti sometimes (e.g. on II 3 21 and V 3 18) 

® He must be older than Bhartrhari, who refers to him in Vakyapadiya II 483, 
and should have been alive when the Huns invaded India because of the example 
ajayaj Jarto Hunan (Cv I 2 81) “Jarta defeated [during the speaker’s lifetime] 
the Huns” [B. Leebich, Ksiratarangini (Breslau, 1930), p. 264^272]; cf. above, p. 

10 Cv 1 1 145 svara-visesam astame vahsydmah. 

11 A Vedic rule is quoted in Cv 1 1 6, accent rules in Cv I 3 45 and V 1 81 ; Candra- 
gomin’s root list contains Vedic roots. 

12 The use of the old pitch accents was, however, not unknown in the oldest 
Buddhist traditions (S. Levi, JA XI® serie, tome V. 401-447). The separate place- 
ment of Vedic and accent rules at the end of the grammar had its impact on later 
grammars; Vedic and accent rules are placed in two appendices in the Siddhanta 
Kaumudi, and Hemacandra and Kramadi^vara who had no use for them gave 
transfer rules for the derivation of the Prakrits instead. 

13 Of the once large secondary literature, we have remainders in Nepalese manu- 
scripts and in Tibetan translations (B. Liebich, GGN 1895. 31 7f., summarized in 
lA 25.103-105). 

1^ W. Rau , ZDMG 113. 521-529. Its onetime popularity did not save the grammar 
when Buddhism vanished in India. In Ceylon it was replaced by a derived popular 
version called Bal5.vabodhana, composed by the monk Ka4yapa c. A.D. 1200. 

13 References in the third person are in themselves not uncommon in autocom- 
mentaries, e. g. Vi^vanatha’s Sahityadarpana I 1 . 

The Buddhist Sanskrit grammarians 


refers to sutra-s of Panini rather than to Candragomin’s parallel sutrai® and 
derives words with unadi-suffixes (e. g. on V 1 9) which are not foxmd in Candra- 
gomin's Unadi-sutra-s. R. Birwe^^ proposes that the author of the vrtti on the 
books I to IV should be Candragomin, of the vrtti on the books V and VI 
Dharmadasa, whose name appears in a note at the end of the only complete 
manuscript in the hand of another scribe: ‘‘This is the work of the venerable 
teacher Dharmadasa.’’ This is not a likely solution. The note with Dharmadasa’s 
name has no separate bearing on the vrtti alone and Birwe’s division of the 
vrtti (for the reason that all nine cross-references in the first person occur in 
the books I to IV, most of those in the third person in books V and VI) is 
arbitrary. One can as well point out that seven out of nine first person forms 
refer to the accent rules of book VIII, while only one out of five third person 
forms refers to the accent rules, but three to closely following rules— the differ- 
ence maybe due to stages in Candragomin’s composition of the vrtti. Chatterji’s 
assumption that the vrtti is not only written by another later author (Dharma- 
dasa) but is even later than the Ka^ika is hardly tenable. The Ka^a on III 
3 175 argues against Candravrtti 13 4 and it is unbelievable that the Candra- 
vrtti on 1 3 106 had to look for the Ka^ika on III 3 131 to find the reason why 
Candragomin had no rule corresponding to Panini HI 3 131-138, especially 
when the Ka^ika teaches these rules and merely refers to the possibility that 
they could be omitted. The dependence of the Kafflsa on Candragomin’s sutra-s 
is a long established fact.^s 

Candragomin’s grammar is the first great remake of Panioi’s grammar (assur- 
ing us incidentally of the reliability of our Panini text) and has had a lasting 
effect on the later Jain revisions of it. Its main claim to originality is staked 
out in the author’s ovni words (Cv II 2 68) Candrdfaj^m a^ainjuaJcam vydka- 
ranam “The termless grammar is the work of Candra.” To an astonishing degree 
he has succeeded in using only enumerations, contractions or descriptive ex- 
pressions and in avoiding defined terms. Wherever Panini would use the term 
vrddhi for vowels of the highest ablaut grade (d [dr, dl] ai, au) Candragomin 
uses the contraction (e.g. V 1 83: augment plus initial root vowel > 

vrddhi). While Panini called a word with a vrddhi vowel in its first syllable 
vrddha (Panini 1 1 73), Candragomin’s formulation remains descriptive: 11 4 98 
d^-ai^-ddy-a<^.o “Suffix -ya after [a word] which has {d . . . ai, au} as the 
first {a . . . au} (i.e. vowel).” The most striking application of this principle is 
in the syntax of cases where Panini’s three levels (above, p. 94-96) have been 
reduced to two: objects and case forms. The expressions Candragomin uses 

It is also curious that Candragomin, who replaced in several sutra-s Panini’s 
sam'jndydm ‘when it is a term’ with ndmni ‘when it is a name’ (e. g. 1 2 30 ; 3 77 ; D 2 
14), paraphrases it in his vrtti with this same samjndydm. 

R. Birw: 6 in Melanges d’Indianisme [Fs. L. Renoxj] (Paris, 1968), p. 127-'142. 
F. KiELHOBiiT, lA 15.183-185; the Ka^ika never openly acknowledges this 
debt by more than a ‘some say . . .’ 


H. Scliarfe • Grammatical Literature 

correspond either to Panini’s objective relations {ddhdra locus’ or hriydpya 
‘attainable by action/ i.e. object) or his semantic concepts (hartr ‘agent/ 
sampradana ‘recipient’)— all of them used descriptively. The semantic concepts 
whose introduction W.D.Whitneydeplored(“thevastlymoredifficultanddange- 
rous method”)!® have disappeared between the two levels .20 Only the Pali 
grammar of Moggallana has followed Candragomin in this reduction. The case 
suffixes including their replacements are taught in II 1 1—42 followed by the 
rules on their syntactic use (11 1 43-98). There is no rule corresponding to 
Panini’s 11 3 1 anabhihite “When [the semantic relation] is not already expressed 
[otherwise]” prefixed to the section. The generative strings which we have in 
Candragomin’s substitution rules for nominal and verbal suffixes do not seem 
to extend to the more fundamental concepts. 

Candragomin’s material contributions to grammatical description are few: 
the wider use of the genitive case (II 1 95), the use of the negation md with the 
imperative and future (13 4: md harotu ‘he shah not do’ ; condemned by Kasika 
III 3 175), the alternative construction of He ‘without’ with the accusative as 
weU as the ablative (II 1 84), etc. Most of these additions have been subse- 
quently introduced into the Paninian system by the Kasika. On the formal side, 
Candragomin’s formulations are often shorter than Panini’s. Of the three 
synonymous expressions vd, vibhdsd, anyatarasydm used by Panini, Candra- 
gomin only uses vd because it is the shortest. Often words are dropped or re- 
placed by new, shorter expressions. Against Panini’s 14 2 vipratisedhe param 
hdryam “When there is conflict, the subsequentpy formulated] rule [takes 
precedence],” Candragomin dittoes parah from I 1 14 into 1116 vipratisedhe 
though here a neuter param is needed, and omits hdryam. The solution was not 
ideal but the idea caught on; the Jainendra-vyakarana uses a shorter synonym: 
I 2 90 sparddhe param “When there is rivalry, the subsequent [prevails]” and 
Sakatayana-vyakarana 1 1 46 sparddhe with parah dittoed from 1 1 44. 

Candragomin adheres to Painni’s metalanguage and develops it in some de- 
tail (e.g. a verbalizing zero suffix VI'p I 1 27^! and a primary suffix as in 
tddrk-sa 1 2 51). This adherence is the reason for his peculiar remodelling of the 
sound table (merging 4^ha ya va raP and 5 laV' into one sutra ha ya va ra laV') 
which did not remove the double use of n as an end marker (1 a % uV' is the other 
occurrence). K.C. Chatter ji wonders in his note on I 1 5: “It is a thousand 
pities Candra did not remove the most glaring defect ...” Such a switch of end 
markers however would have meant a change in the metalinguistic formulae 
while the elimination of the end marker t amounts only to non-usage of an 
already unnecessary term {aP occurs four times in the Astadhyayi but is not 
needed by Candragomin). 

W.D. Whitnev, AJPh 14. 171. 

Or shall we assume that Candragomin dropped the objective relations xmder 
the influence of Buddhist mentalistic philosophy ? 

Of. Mahabha§ya II 21, 16 f. 

The Buddhist Sanskrit grammarians 


The topical arrangement of the grammar (primary noun suffixes, conjuga- 
tion, declination, secondary noun suffixes, morphophonemic rules) stays 
closer to Panini’s arrangement than we find it in other grammars with 

topical arrangements.22 

22 The two Tibetan grammars (before the 10th cent. A.B.) which tradition ascri- 
bes to Thonmi Sambhota [Les 61okas grammaticaux de Thonmi Sambhota, ed. 
J. Bacot (Paris, 1928)] show a very general influence of Sanskrit grammar ; cf. 
S. IrTABA, JIBS 3.432-440 and R.A.Milleb, JAOS 83.485-502 and ZDMG 115. 
327-340. In China Sanskrit grammar was not studied seriously; for a summary 
description of the Sanskrit language, see the reports of the pilgrims Hsiian-tsang, 
I-tsing and Pa-tsang in J.P. Staal, A reader, p. 7-19. It must be regarded as an 
unusual achievement that the Japanese monk Jiun Sonja Onk5 inductively ab- 
stracted (sometime between A.D. 1751 and 1771) the rules of Sanstet ^ammar 
from Buddhist texts that were available in manuscripts ; his work exists in manu- 
script form in the Kokiji temple, only a part of it is published pR.H. van Gmm, 
Siddham (Nagpur, 1956), p. 133-135 and Watanabb Shoko, Japanese Buddhism 
(Tokyo, 1970), p. 24]. 



In time the Jains too adopted Sanskrit as a vehicle of thought and wrote 
grammars to suit their needs. Rules teaching purely Vedic forms and the old 
pitch accents were dropped, but neither the Jain grammarians nor any other 
Indian grammarians have replaced the latter with rules on the often-assumed 
stress accents in Sanskrit or other Indian languages. The reason for this omis- 
sion is possibly that the stress accents, if they exist, are not strong.^ 

The oldest work is the Jainendra-vyakarana of Devanandin called also 
Piijyapada. While some scholars place Devanandin before even Candragomin 
(oth cent. A. D.), others put him later than the authors of the Kasika (early 7th 
cent. A.D.).2 The grammar follows strictly Panini’s order of rules and retains 
their generative character. It seeks to make its contribution in the refinement 
of details, 3 especially in the further economy of expression. Prom the word 
vihhakti 'case' Devanandin forms a variant v4-bh-a-k4-l to denote the seven 
cases by adding d to the consonants and p to the vowels (I 2 158): w = nom.,,, etc.^ The legion of invented short terms taxes the memory of the 
student. Devanandin follows the suggestion of Katyayana that the 'single 
remainder' process (above, p. 136) is not necessary to account for the dual and 
plural forms, and his commentators Abhayanandin® and Somadeva® refer to 
the work as the 'grammar without single remainder' — even though in Soma- 
deva’s recension the single remainder process has been reinstated. This would 
he an unhappy characterization if Devanandin lived after Candragomin who has 
also e limin ated the single remainder process. Devanandin stays so close to 
Panini’s formulations that some additions proposed by Katyayana or Patanjali 

^ Cf. SuNiTi Kukab Chattebji, IL 21.81f, 

2 p,. Brawn’s Introduction, p. 40 f., to Sambhuntath Tbipathi’s edition of the 
Sakatayana-vyakarana (Benares, 1971). 

® A happy formulation is I 4 54 mi^.aihdrthe vdh “nominative suffix: for that in 
congruence with the verbal ending” against Panini’s H 3 46 (above, p. 95f.). JSTote, 
however, the reservations voiced by Katyayana 11 3 46 vartt. 6. 

^ Alone of all Indian grammarians Devanandin begins the list of verbal personal 
suffixes with the first person followed by the second and third. The reason is expe- 
diency which allows him to contract Paimi’s rules 1 4 105 through 108 into a single 
rule I 2 152; now. the three verbal suffixes are used parallel to ‘we,’ ‘you’ and ‘the 

5 In his Mahavrtti on I 4 97 and HI 3 84. 

® In his Sabdarnava-candrika on I 4 114 and III 3 98. 

The Jain Sanskrit grammarians 


appear only in Abhayanandin’s commentary. Against the 3063 sutra-s of the 
northern (and original) recension, the southern recension as commented on by 
Somadeva has — due to numerous alterations and additions— 3708 sutra-s. 

In the ninth century A.D., a Jain monk, whose real name was perhaps 
Palyakirti, composed his Sakatayana-vyakarana (using the pen name ^akata- 
yana after the famed forerunner of Panini) and a commentary on it named 
Amogha-vrtti, after his patron the Rastrakuta king Amoghavarsa I. He uses 
all preceding grammars, especially the southern recension of the Jainendra- 
vyakarana, though without its excessive use of artificial terms. The topical 
arrangement of the rules reminds one of the Katantra ; the generative character 
is lost and the kdraka-s have a mere shadowy existence. The instruction on 
gender (in 69 dryd verses) is included in the commentary on 1 2 1, and the roots 
are (with the unexplained exception of those belonging to the first class) listed 
in the commentary to several sutra-s of the fourth book where the sutra refers 
to a certain root class. 

The Jain polymath Hemacandra Suri ( 1089 - 1172)7 composed his Siddhahai- 
macandra at the request of his patron, the Calukya king Jayasimha-Sid- 
dharaja of Gujerat, to rival King Bhoja’s Sarasvati-kanthabharana, mainly on 
the basis of the ^akatayana-vyakarana.^ Its c. 4500 sutra-s are organized in 
eight books of four chapters each; but nearly a quarter of them belong to the 
eighth book dealing with the Prakrit dialects which are derived horn Sanskrit 
by means of transfer rules. Hemacandra also wrote an extensive commentary 
on his work, the Brhadvrtti (with a subcommentary, the Brhannyasa, valuable 
for its identification of sources) , and an abbreviated version of it, the Laghuvrtti- 
The Brhadvrtti includes not only the gana-s but also the irregular formations 
(u^-ddi), the gender rules (after sutra 1 1 29, in verses) and a collection of me- 
tarules (called nydya, at the end). In comparison to Bhoja's grammar where 
these lists are incorporated in the sutra-s, Hemacandra’s grammar is conserva- 
tive.® It eclipsed, due to its clear design, the grammars of Devanandin and 
Sakatayana within the Jain communities and had, like these, its share of com- 
mentaries and recasts. But Panini’s grammar and the Mahabhasya were ob- 
viously well studied by all Jain grammarians^® and were never replaced as the 
classical authors on the subject. 

7 G. Buhuee, tlber das Leben des Jaina Monches Hemachandra (Wien, 1889; 
DAWW 37). 

® Still, he claims some originality when he calls his work in the colophons sv6- 
pa/na ‘self-devised.* 

® Curious peculiarities are the replacement of eva by iva in unstressed positions 
{iha-\'eva>i}ii>va instead of ihaiva; 12 16) and the optional pronominal forms 
asahhyam and yusahhyam, besides asmdbhyam and yusmahhyam (II 19). 

Malayaghi’s Sabdanui^asana (ed. B.J. Dosm; Ahmedabad,^ 1967) is ^ un- 
inspired compilation written in the late 12th cent., based on Sal^tayana ^d Hema- 
candra. The grammar of Buddhisagara Suri, called Buddhisagara-vyakarana or 
Pancagranthi (written in A.D. 1027) is still to be edited (B.J. Doshi, Introduction, 
p. 2, fn. 2). 

Chapter XI 


The first author after Patanjali whose work we still have is Bhartrhaii (a 
disciple of Vasiirata, the adversary of Vasubandhu), who is now believed to 
have lived c. A. D. 450-510.^ The information of the Chinese pilgrim I-tsing that 
Bhartrhari had died 40 years before his arrival in India (A.D. 671/672), i.e. in 
A.D. 631/632, must be wrong because Dignaga (c. A.D. 480-540) and other 
Buddhist authors of his time quote and use Bhartrhari’s Vakyapadiya.^ Besides 
this meagre biographical information there are numerous legends, most of them 
based on his supposed identity with the poet Bhartrhari. 

Bhartrhari’s commentary on the Mahabhasya has survived in a single manu- 
script, incomplete and corrupt.^ The very beginning is missing and the frag- 
ment ends with the comment on Panini 1 1 55 though the work is known to have 
covered at least the first three The colophons call it either Bhartrhari’s 

Mahabhasya-tika^ or Bhartrhari’s Mahabhasya-dipika;^ the parallel with the 
Trikandi suggested further the name Tripadi 'The Three Chapters. ’7 It is a very 
learned commentary that records Vedic usage and various opinions of other 

Bhartrhari’s fame mainly rests on his Vakyapadiya 'Work Dealing with 
Sentences and Words,’ the major part of which has survived precariously to 
this day. The first two books (handa) with 1388 and 487 sloka-^, together with 

1 E, Fbatjwallner, WZKSOA 5.134f. 

2 H. Kajbzamtoa, Studies in Indology and Buddhology, [Es. Susitmtj Yamaguchi] 
(Kyoto, 1955), p. 122-136; E. Fbattwaulneb, WZKSOA 3. 107-114. 

/io commentary was still known to Nagojibhatta and his disciple Payagunde 
{18m cent.) who both quote from it. Largely inspired by it is the Mahabhasya- 
pradipa of the Kashmiri Kaiyata (probably 11th cent. A.D., certainly before 1150: 
V. Raghavan, job 19. 223). 

t ^ ^he^ti on VP I 82, Bhartrhari refers to his ( ?) commentary on the Maha- 
bha^ya on Panini I 4 109: that could suggest that he commented on more than the 
three pdda-s. 

5 IWs earned Bhartrhari the name tlha-kara (KumMa, Tantra-varttika ed. 
Ohowkhamba Sanskrit Series, p. 207). 

® V. SwAMmATHAn, ALB 27. 60 

! ALB 35. 159-171 argues that this was the original title. 

Ihe number is slightly higher in some editions depending on the recognition of 
^me verses that may really be quotations in the author’s own commentary taken 
from earlier works (A. Aklujkae, JAOS 91. 510-513). 

The later Panini school 


the author’s own commentary (vrtti)^ are properly caUed Vakyapadiya; the 
third book called Prakirnaka ‘Miscellany’ (with 1320 MoJca-s in 14 sectiom; at 
least two more sections are known to be lostio) treats special points that were 
only outlined in the first two books. Together the three books could be called 
Trikandi ‘Three Books’ but often the name Vakyapadiya was extended to 
cover the whole work.^i 

The relative chronology of these works (not to mention the lost Sabda-dhatu- 
samiksa) is not yet settled, except that the Mahabhasya-dipika (ABOBI ed., 
p. 46, If.) quotes Vakyapadiya II 464. There are many correspondences 
in the formulations of these works, though the vrtti with its long, involved 
sentences (reasons and qualifications piled one upon the other, their attributes 
trailing) is especially hard to understand .12 The Dipika (p. 32, 24 vydharam- 
granthesu lihhitdh ‘written in the grammar books’ ; 33, 17 granthesu cdlikhitatvdt) 
shares an interest in writing not otherwise found in Paniniya works with VP 
1 20 (yatra vdco nimittdni cihndnwdksara-smrteh ‘that in which the manifestors 
of speech like the signs of the alphabet appear . . .’) and the vrtti on I 23 
[aksara-nimittdksara-kalfandvat ‘as the letters of the script are thought of as the 
phonemes of the alphabet ...’). 

The Kashmiri Helaraja (10th cent. A.D.) wrote separate commentaries on 
the verses of all three books of the Vakyapadiya (called Sabda-prabha, *Vakya- 
pradipa? and Prakirnaka-prakasa) of which the first is lost and the second 
wrongly attributed to Punyarajai^ in most manuscripts two gaps in the 
Prakirnaka-praka^a (III 7 34-49 and 65-69) have been filled with the comments 
of one Phullaraja. Vrsabhadeva (date unknown) further wrote a ^os^(pMhati ) 
on the first kdnda, including the vrtti. 

Ehartrhari asserts the value of traditional interpretation and criticizes (VP 
11481) Baiji, Saubhava and Haryaksa for following ‘dry reason’ (suska4arka^^), 
i. e. following factual observations of their own rather than the intimations in 

® The vrtti is an integral part of the work. Several times a sentence combines 
portions of the commentary and a verse (e.g. I 65; 90). Often the verses are formu- 
lated so that the vrtti can fill them with alternate interpretations to suit linguists of 
different persuasions (A. Aklujkajr, WZKSA 16.181-198; esp. p. 186). Only the 
vrtti has clear references to the illusionist concept of evolution which led M. BiAit- 
BEAU to contest Bhartrhari’s authorship of it [Bhartrhari-Vakyapadiya-Brahma- 
kanda avec la Vrtti de Harivr^abha (Paris, 1964), p. 5-21]. 

Helaraja’s (?) commentary on VP II 77. 

A. Akbiukab, jags 89.547-554. Sometimes the three books are referred to 
as Brahma-kanda, Vakya-kanda and Pada-kanda. 

12 Detailed characterization by K.A. Sttbbamania Iyer, The Vakyapadiya of 
Bhartrhari with the vrtti, chapter I, English translation (Poona, 1965), p. xii-xvi. 

Punyaraja (probably later than Helaraja) is the author of a verse summary of 
Kanda II. 

A. Aklujkar in Charu Deva Shastri Felicitation Volume (Delhi, 1974), 
p. 165-188. 

Defined in the vrtti on VP 1 129. 


H. Scharfe ' Grammatical Literature 

works like tke Mahabhasya. Agreeing with Patanjali (Mbhas. I 1, 19), Bhar- 
trhari (VP I 11) regards grammar as the most important veddnga\ for him 
though, its value is not so much in its leading the adept to prosperity and to 
heaven (through conscious correct usage^®) as in its revelation of brahman 
(through meditation exercises centered on language: vdg-yoga or sabda-purva- 
yoga)y^ Speech is brahman. This central role of speech goes back to early Vedic 
tradition and was stressed again in tantric systems with which Bhartrhari 
shares a few key terms ( sakti, sddhana) ; it is not surprising that Abhinavagupta 
of the Kash m irian tantric school pays tribute to Bhartrhari’s work.i® Speech and 
meaning coexist in an undifferentiated state, from which the diversity of 
objects unfolds due to ignorance (avidyd) through a wrongful attribution of 
differentiation by time and space. To convey such objects, comprehensive 
language signs develop, existent only in the mind and without inner sequence. 
To denote them, the ancient word sphota ‘burst’ is pressed into service, which 
in Mbhas. I 181, 21 denoted the permanent aspect of a phoneme.^® On the 
phonetic level, a sphota is then manifested by sequential sounds (dhvani), 
forming a sentence.2« From this sentence the listener realizes the sphota within 
himself and understands the communication. For the grammarian to analyse 
the communication or the sentences, it becomes necessary to superimpose for- 
mal categories like words, morphemes and phonemes/sounds which he would 
even impute to the sphota; thus we can speak secondarily not only of a sentence- 
sphota but also of a word-spAoto, ^oxmd-sphota, etc. 

Already Katyayana recognized the importance of the ‘desire to express’ 
( vivaksd) in language, but only Bhartrhari has given it a philosophical founda- 
tion. Besides ‘primary Being’ [mukhyd sattd VP III 3 46), he recognizes ‘meto- 
nymic Being (aupacariki sattd VP III 3 39), which is the meaning of words and 
exists in the mind. This new concept offers a better solution to an old problem: 
how can we say “Weave a cloth!” using the word ‘cloth’ for a thing that does 
not yet exist (if it already existed there would be no need to make it) ? Patanjali 

Incorrect forms may serve the secular purpose (below, p. 191) but no merit 
( aharma) accrues from their use. 

17 K.A. SUBRAMANIA Iyer, ALB 28.112-131. In the Dipika, p. 33,24 - 34,1, 
Bhartrhari argues that, due to the permanent connection of word and object, words 
mce apurva ‘fate’ let us infer the existence of their invisible objects. That may, 
however not be his own conviction, for in VP III 3, 39-51 he recognizes that the 
object of a word is a mental image, not an external object. 

1® A. Padoux, Recherches sur la symbolique et I’dnergie de la parole dans certains 
textes tantriques (Paris, 1963), p. 16. Abhinavagupta wrote a commentary on the 
tnira kdnda called Prakimaka-vivarana. 

1 123 mentions a grammarian called Sphotayana; any connection 
with the later sphota theory is speculative. On Audumbarayana’s theory, cf. above, 


^ In VP 1 134/142 we hear of three levels of speech (vac) which are given as 
pa^anK ( seeing:’ undifferentiated), madhyama (‘middle:’ sequential and mental) 
and vaikhari (‘elaborated:’ spoken language). 

The later Panini school 


answered that ‘cloth^ is a ‘potential term* (bhdvini samjnd) denoting that which 
^ be called a cloth once it is completed (Mbhas. 1 112, 9-13; above, p. 161). 
Bhartrhari’s concept of meaning as a mental image covers not only such 
instances of creation but also negated and absurd notions i the existence of a 
meaning does not imply reference to an external object. 

Pardni’s system of kdraha-s receives by the same device a proper philosophi- 
cal underpinning. The naive impression that the hdraha-s are relations, that 
Icarrmn ‘object* or harana ‘instrument* are things participating in an action, 
must give way to the insight that we deal on the semantic level not with things 
but oiy with the inherent powers of things to contribute to the action and that 
the mind exercises a great deal of freedom in the selection of the powers it 
wishes to express (VP III 7, 1-3). The simple fact that Devadatta cooks rice in 
a pot is most commonly visualized with Devadatta as the agent and the pot as 
the location (Devadattdh stJidlydm odanam ^pacati); if however the speaker 
wants to stress that due to the quality of the pot the rice cooks quicker he 
may denote the pot as the agent {sthdli pacati “The pot cooks’*). Similarly the 
firewood (normally seen as an instrument) or the rice itself may be visualized 
as the agent, but usually not a ‘recipient* (sampraddna) or a ‘take-off* fapd- 
ddna), i.e. there is no way to visualize Rdma in Edmdya dadati (“He gives to 
Kama”) or vrksa in vrksdt patati (“He falls from the tree”) as an agent. Rdmo 
dadati “Rama gives’* or vrksah patati “The tree falls” would describe totally 
different events.^i Such changes of concept and expression are possible because 
each kdraka is an agent of its own contribution to the main action and assumes 
the role of instrument, etc. only in relation to the main agent (VP III 7, 20-23). 
Following the lead of Patanjalij^^ Bhartrhari differentiates various services 
(upakdra) of the kdraka-^, e.g. three major and four minor ones for karman 
‘object* (HI 7, 45) because ‘object’ may be the product {mrdd ghatam karoti 
“He makes a jar with clay*’) or a modification [mrdam ghatam karoti “He makes 
the clay into a jar”) or destination (nagaram upasarpati “He goes to the city”). 
The four minor classes extend the object notion from ‘that most desired by the 
agent* to unintended things (‘walking to the city he touches grass’), disliked 
things (‘he eats poison*), the residuals of Panini I 4 51 (the double acc. with 
verbs meaning ‘ask,* etc.) and those cases where an added preposition demands 
the replacement of another MmA;a-and-case(]//:md^ with dative, but abhi ^krudh 
with accusative). This line of research must have been continued by later 
authors because the Tamil grammar Viracdliyam refers to an elaborate system 
of upakdraka-^ within the range of the kdraka-s (below, p. 182). 

VP III 7 18 with Helaraja’s commentary. Already Katyayana stated in his 
varttika-s 7 through 14 on Panini I 4 23 that ‘instrument’ (harana) and ‘location’ 
( adhikarana ) can be visualized as ‘agent’ (kartr ) , but ‘take-off’ ( apdddna ) , etc. can not. 

22 Mbha§. HI 51, 8f. “Location is of three kinds: pervasive, touching, topical.” 
The formulation suggests that Patanjali quotes another source [H. Schaufe, Die 
Logik im Mahabha§ya (Berlin, 1961), p. 76]. 


H. Scharfe • Grammatical Literature 

The study of Bhartrhari's thought is still in its infancy; critical editions and 
usable translations come forth only slowly. But the study should be well worth 
the while of the linguist. 

Though Panini’s sutra-s must have been accompanied by oral instruction 
that turned the sutra-s into understandable sentences and supplied the student 
with sets of examples, no such old vrttihas survived. The Kasika- vrfcti of J ayaditya 
and Vamana (early 7th cent. A.D.),23 in spite of its title ‘Vrtti from Benares/ 
wants to be more: a compendium of Paninian grammar. Its pedagogical 
approach has made it one of the most popular grammatical works, but its 
simplicity is often deceptive^^ and the full meaning of its remarks becomes 
apparent only against the backgroimd of earlier discussions, notably in the 
Mahabhasya. The commentary follows a sutra text that often adopts the altera- 
tions proposed in the Mahabhasya.^s The Buddhist Jinendrabuddhi (8th or 9th 
cent. A.D.) wrote the voluminous commentary Kasika- vivarana-panjika alias 
Nyasa on it in which he demonstrates an interpretation of the Astadhyayi that 
makes Katyayana’s varttika-s redundant; Haradatta’s Padamanjari (13th cent. 
A.D.?) summarizes the Kasika philology of his time. The Bhasa-vrtti of the 
Buddhist Purusottama (early 12th cent. A.D.) further simplifies the presenta- 
tion of Paninian grammar and omits the rules concerning Vedic language. 

Whereas the Astadhyayi is a scientific analysis for the benefit of native 
Sanskrit speakers, classroom needs called for a more practical introduction into 
Sanskrit. It is hardly an accident that the first such work within the tramework 
of the Paniniya tradition was produced by a Buddhist author from Ceylon: 
Dharmakirtfs Rupavatara (10th cent. A.D.?) teaches Sanskrit in the form of a 
catechism arranged by grammatical topics similar to the Katantra and often 
illustrated by paradigms. Other such rearrangements of Panini’s sutra-s are the 
Rupa-mala of Vimala-sarasvati (14th cent. A.D.), Prakriya-kaumudi of Rama- 
candra (14th or 15th cent. A.D.), Praknya-sarvasva of Narayanabhatta (A. D. 
1616)26 and the Siddhanta-kaumudi of Bhattoji-diksita (early 17th cent. A.D.). 
The last work especially has become the basis of several commentaries (two by 
the author himself: the Bala-manorama for students and the Prau^a-mano- 
rama for scholars) and has all but replaced Panini’s work and the Mahabhasya 

^23 JayMitya is the author of books I to V, and after his death in A.D. 661 
Vamana completed the work. For an occasional difference of opinion between the 
two authors, cf. Praudha-manorama on Panini V 4 42 and Y. Ojihaba, JIBS 9. 
749-753; 766-776; 845-847. 

24 ';]^0 examples for the application of Panini ’s rules are often taken from discus- 
sions in the Mahabhasya where they illustrate marginal applications. Many other 
examples are taken from Candragomin’s vrtti. 

25 F. Kielhorn, IA 16. 178 and KGGW 1885. 190. 

K. Kunj ujstnt Raja, The contribution of Kerala to Sanskrit literature (Ma- 
dras, 1958), p. 129. InKrt-khanda p. 82 = VTI3 62 Narayanabhatta recognizes the us- 
great authors like Murari and Bhavabhuti as authoritative. He also wrote the 
Dhatu-kayya which narrates the E:r§na legend using all 1948 roots found in Bhima- 
sena’s Dhatupatha — even in the same sequence! 

The later Panini School 


for many students of Paniniya Sanskrit grammar. The curious fact is that these 
sutra-s, being identical with those of the Astadhyayi except for their sequence, 
depend on a knowledge of Panini’s original work for their interpretation (e.g. 
the dittoing of words from a previous rule which the author supplies in hL 
paraphrases). 27 The arrangement (developed by Sarvavarman, Dharmakirti 
and Ramacandra) proceeds from vowel and consonant sandhi to the inflection 
of masculine noun stems ending in vowels, etc., various compounds, secondary 
noun formation, verb inflection and primary noun formation.^s It deserves to 
be studied from a linguistic point of view in contrast with the previous rearran- 
gements and with Panini’s generative strings. Because so much in grammar is 
formal, it is not a meaningful critique of these works to say that “they differ 
only in the arrangement of the material.’’ 

It is not possible here even to mention the many authors that contributed 
through commentaries or monographs to the later Paniniya tradition; many of 
their works still exist only in manuscripts. They are aU put in the shade by the 
Mahratta Brahmin Nagesa or Nagojibhatta Kale^^ (died in 1755 in Benares), 
w^hose prolific writings cover not only grammar but also dharma, poetics and 
yoga. His Mahabhasya-pradipoddyota is a subcommentary on Kaiyata’s com- 
mentary on the Mahabhasya; his Brhacchabdendusekhara, a subcommentary 
on Bhattoji-diksita’s Praudhamanorama. An independent work is his Vaiya- 
karana-siddhanta-manjusa (in three recensions of different length), a work of 
great depth in which he carries on the work of Bhartrhari and the tantric 
philosophers-20 One of his last works may have been the Paribhasendusekhara 
‘Moon crest of metarules’ in which he critically examines previous attempts to 
gather all metarules applicable to Panini’s grammar. He condemns the search 
for even more metarules beyond those recognized in the Mahabhasya and advo- 
cates strong reliance on Patanjali as the latest authoritative source. 

27 Actually Bhattoji-diksita and his followers led a strong revival of Paninian 
studies at the expense of the non-Paninian systems. 

28 A late echo is M.R. Kale’s A ffigher Sanskrit Grammar (Bombay, 1894). 

28 He was a disciple of Bhattoji-dik§ita’s grandson, Hari-dik§ita. 

D. Sbyfort Rtjegg, Contributions a Thistoire de la philosophie iinguistique 
inclienne (Paris, 1959), p. 5-14. The manuscript of the Brhanmanju^a, kept in the 
library of the Government Sanskrit College, Benares, is probably written in the 
author’s own hand [GornsrATH KAvxeiaj in the foreword to Pt. Sabhapati’s edition 
of the Laghumanju§a (Benares, 1963), p. 1, fii. 1]. The saMa-bodha philosophy of 
Naiyayika-s, etc. has lately attracted the attention of some modern scholars. To 
mention only one point, action (kriyd) is thought to consist of operation (vydpdra) 
and fruit (phala) of which the former resides in the agent, the latter in the object; 
if both coincide we have an intransitive verb [K. Madhava KAishka Sarma, 
Panini Katyayana and Patanjali (Delhi, 1968), p. 160-164]. The contributions to 
linguistics contained in the works of the Mimamsa and of poetics cannot, in this 
limited survey, receive the attention they deserve. 

Chatter XTI 

THE glK^A-S 

The many works regarded as Siksa-s^ cover a wide spectrum of Vedic studies. 
The dating of most of these texts is next to impossible ; while some Siksa-s, e. g. , 
the Narada-^iksa or the Apisali-^iksa, may go back to the 5th century A. D., if 
not further, 2 others are much younger and must be assigned to the 11th to 15th 
centuries.3 In spite of names like Apisali-siksa or Panimya-siksa, none of them 
preserve the doctrines of the pre-Paninian or Parunian era. In the Panimya- 
^iksa both /r/ and /r/ are called retroflex sounds; but according to the Prati- 
^khya-s /r/ was pronounced at the root of the tongue and /r/ at the roots of the 
upper teeth. We would be mistaken if we accepted this Siksa as a witness for 
original Vedic pronimciation; it can only reflect the late tradition of a medieval 
school. The real thrust of the ^iksa-s is more elementary; it is revealed for 
example by a little treatise called Svara-vyanjana-siksa which has only one 
purpose: to teach the student when he should regard an r sound found in a 
text as a vowel and when as a consonant. To understand the problem one has 
to remember that the vocalic /r/ of Sanskrit did not survive as such in the suc- 
cessor languages, and its pronunciation in academic Sanskrit was more Kke /ri/: 
Pig- Veda, Samskrit, etc. It thus became mdiscemible from original /ri/ sounds: 
risddas, rip% 'pratarindram (i.e. prdtar Indram), and special rules became 
necessary to guide the student in his pronunciation and spelling.^ This concern 
with orthography is also clear in a quotation found in the Gautami-siksa which 
deals with the common doubling in consonant clusters: the Siksa quotes from 
another text the form yunnhshsva^ (instead of the normal yunJcsva). The form 
looks bizarre until we consider the aksara ks as a graphic unit — ^the proper 
transcription would be yunnhkssva. 

Almost all Siksa-s are attached to a certain Veda; the Taittiriya school of the 
Black Yajurveda was the most prolific. It is noticeable that a great percentage 
of these manuals are products of South India; the Pari-^iksa by Cakra is even 
dedicated to the memory of Chief Pari, praised in the Old Tamil literature. 

1 SmDHESHWAR Vabma, Critical studies in the phonetic observations of Indian 
grammarians (London 1929, repr. ed. Delhi, 1961), p. 29, knows of sixty -five. 

2 B.A. VAN XooTEN, Oriental studies 2 (Tartu, 1973), p. 408-438, considers the 
possibility that the Apisali-sik§a is even older than the Mahabha§ya. 

2 SroDHESHWAR Vabma, Critical studies, p. 28-52. 

4 Seddkeshwar Varma, Critical studies, p. 58f. 

5 SroDHESHWAR Varma, Critical studies, p. 51. 

The Siksa-s 


When Vyasa-siksa 317 defines the duration of one measure (i.e. the duration 
of a short vowel) by comparing it with the ‘snappiug of the fingers’ (angvli- 
sphotana), it is infiuenced by the Tamil grammarians (TolkappiyamI 7, Nannul 
100: noti). The Vyasa-siksa (13th cent. A.D.) develops the phonetic observa- 
tions contained in the Pratisakhya of its school, i. e. the Taittiriya-Pratisakhya, 
especially the aspects of quantity. The duration of the nasal sounds in various 
contexts is stated minutely — ^not to speak of the different pauses in a hiatus, 
after a sentence, a verse or a half-verse, at the end of a section or a chapter. On 
the other hand, numerous sandhi rules raise it almost to the rank of a Prati- 
^khya. A curious feature of this and many other Siksa-s is the symbolic value 
attributed to sounds that somehow connects sounds with castes and certain 
tutelary deities. The Mandavi-siksa of the White Yajurveda probably originated 
in Central or Northeastern India, where the distinction of /b/ and /v/ was lost, 
for it consists of an enumeration of the 641 words in the Yajurveda with ‘true’ 
/b/ sound; similarly, the Amoghanandini-siksa gives a list of words with the 
initial /b/ and another with the initial /v/.® 

® SiDDHESHWAR Vaeima, Critical studies, p. 33f.; 130. 



Of all the Dravidian tongues, Tamil has the oldest attested literature. More 
than 2000 poems, many of them from the 1st to the 3rd century A.D., are 
preserved in the Eight Anthologies that make up the Sangam literature, so 
called after the Tamil Academy or canJcam supposedly held at Madurai.^ Tamil 
also has the oldest grammatical-literary compendium, the Tolkappiyam, which, 
in about 1610 verse-sutra-s^ (in 3 sections: eluttu ‘sound/letter,’ col ‘word,’ porul 
‘subject matter’), comprises all literary activity. It is still an unsolved problem 
whether the Tolkappiyam is older or yotinger than the Anthologies, While 
literature must precede literary theory, the theories found here could refer to 
older, now lost works. Tolkappiyam’s classification of genres forms the basis 
for the arrangement of the Anthologies: was it an ancient tradition or was it 
derived from an observation of the poetry collected for the anthologies? There 
are minor grammatical discrepancies between the Tolkappiyam and these 
poems: the initial /y/ (e.g. ydru ‘river’) is preserved in the Tolkappiyam but 
lost in the poems (dru ) ; Tolkappiyam IE 221 prohibits the use of the first and 
second person optative — ^but such forms are found in the poems.^ If oral (III 
133) is the astrological ‘hour’ (from the Greek mpa cf. Sanskrit hora),^ the earliest 
possible date for the Tolkappiyam would be the 2nd century B. C.® 

The original title of the work and the name of its author remain unknown. 
Following more recent fashion, the title Tolkappiyam is formed from Tolkappi- 
yan, which is given in Panamparanar’s preface (late Sangam period?) as the 
author’s name or, rather, surname: Tolkdppiyan eim ttan peyar tdrri “having 
shown his name to be Tolkappiyan (‘One who [knows] old poetry’).”® The same 

1 K. ZvELEBiL, nj 15. 109-135; Tamil literature, in voL X of this History, p. 7. 

2 The number varies very slightly with the commentaries. The meter is nurpa. 
With the metrical form goes a certain prolixity and rotmdabout expression (H 205 
‘four fives plus three on top’ instead of ‘23’). 

3 C. and H. Jesudasan, A history of Tamil literature (Calcutta, 1961), p. 3f.; 
T.P. Meekakshisxjndaban-, a history of Tamil language (Poona, 1965), p. 51. 

4 Disputed by S. Iiaxettvanae, Tholkappiyam (in English) with critical studies 
(Madurai, 1963), p. lOf. 

5 J. Ftlliozat in L. Renoij and J. Fhtjozat, L’Inde classique, vol. 11 (Paris, 
1953), p. 193. 

® The word happiyam (Sanskrit kdvya) is attested to by Manimekalai 19, 80 etc. 
Much less convincing is the explanation as “Old Kappiyan” in spite of the occur- 
rence of a family name Kappiyan and a potential parallel in Tolkapilar “Old Kapi- 
lar” ; the context demands a descriptive name. 

Grammars of the Dravidian languages 


preface declares him to be full of aintiram which the commentators take as a 
reference to the Aindra-grammar. But, as the Aindra^grammar is only a late 
myth (10th cent. A. D. 1), it is tempting to accept the proposal of K. Subrama- 
nia Phlai,*^ namely, to amend the text to read ain4iram ‘the five sections’ which 
fits the work well even if there are only three formal sections: metrics and 
poetics are treated in the section on subject matter. Panamparanar’s statement 
that Tolkappiyan practiced padimd and Tolkappiyan’s own classification of 
beings by the number of their sense organs (III 571-577) suggest that he was 
a Jain.8 The popular association with Agastya, his supposed teacher, is first 
mentioned centuries later and deserves no credit.^ Still, it is certain that Tol- 
kappiyan had predecessors as his frequent statements “so they say” “so say 
the wise” indicate. Panamparanar says the work was presented at the court of 
the Pantiya king Nilantaru Tiruvil in the presence of the scholar Atankottacan 
‘the teacher from Atahkotu’ (a village in southern Travancore). 

It is not quite clear whether an apparent unevenness (gaps and duplications 
in the grammatical description,^® unexpected placement of some rules^i) should 
be blamed on the lax structure of the composition or on a faulty tradition of 
the text. Ignorance of a rule found in our text^^ "by the author of Yapparunkala 
(10th cent. A. D. 1 ) may be due to a lapse of memory or to a later interpolation.^^ 
Older forms like kalapu appear in our manuscripts as kalavu, etcM The oldest 
commentary is by Ilampuranar (before 11th cent. A.D.?) who seems to have 
started a re"vival of Tolkappiyam studies, followed by Teyvaccilaiyar (on ‘word’ 
only), Peraciriyar, Cenavaraiyar (on ‘word’ only) and finally Naccinarkiniyar 
(14th cent.). The Tolkappiyam in recent times is regarded as the main authority 
on ‘Good Tamil’ (cenramil), rivaled only by Nannul, and largely accomts for 

7 Quoted by N. Stjbbahmanian, Pre-Pallavan Tamil Index (Madras, 1966), 
p. 176; /r/ and /r/ have merged in. some Tamil dialects. It is not clear how the above 
statement on the Aindra grammar would be affected by a quotation foxmd in 
I-hsing’s Chinese commentary on Mahavairocana-sutra, chapter I (early 8th c. 
A.D.), from a ‘vyakarana of Sakra’; after all, Sakra is Indra (Wilhelm Kuho 
Mulleb, Shingon-Mysticism; Subhakarasimha and I-hsing’s conunentary to the 
Mahavairocana-sutra, chapter one, Doctoral Dissertation, Los Angeles 1976, p. 9 
with fn. 39 and 40). 

® The term ticai-ccol ‘dialectal word’ (Pall and Prakrits died) could point to a 
Prakritic source of terminology, expected for a Jain. 

® K. ZvBLEBiL, UJ 15.124, fn. 61, tentatively accepts as genuine the 53 lines 
quoted in later commentaries from ‘Akattiyam.’ 

^0 T.P. MEENAKSHisxjmDABAN', A history, p. 52 f. 

G. Vijayaventjgofal, a modem evaluation of Nannul (Annamalainagar, 
1968),p. 119f. 

12 G. Vijayaventtgopax, a modern evaluation, p. 18f., 119f. 

12 Generally suspect are attempts like those of Llaekuvaitab, Tholkappiyam, p. 
15-'20, to eliminate as interpolations all sutra-s with traces of Sanskrit influence in 
order to claim a high antiquity (700 B.C.) for the Tolkappiyam. 

1^ G. VijAYAVENTjGOPAL, A modem evaluation, p. 155. 


H. Scharfe • Grammatical Literature 

the surprisiiig continuity of literary Tamil as used even now in novels or the 
news media. 

The phonetic section shows the influence of the It is worth 
noting that the ‘retroflex’ consonants /t/ and /n/ are not retroflex in the 
Tolkappiyam, though they are in later Tamil works (and in the Pratisakhya-s). 
The sound system (mixed up with graphemic representation^®) is phonemic; 
there are doubts concerning only some of the five n soTmds. Also, there are 
three dependent soimds or allophones: shorter shorter u and dytam. The 
exact character of the last, whether it was a glottal stop/catch, a diacritical 
mark indicating voicing, or a fiicative,^'^ is still not clear despite a large body 
of literature on the sub j ect. Tolkappiyan has developed an inquiry, first put forth 
in the Pratisakhya-s (Vajasaneyi Pr. I 85-89; Pgveda Pr. XII 1; Caturadhya- 
yika 1 3-7), into which sounds may stand in initial or final position and into 
which combination of consonants may occur generally. He rightly exempts 
metalinguistic references from these rules (1 47 + 66). Tolkappiyan does not refer 
to the non-phonemic voicing of intervocalic simple consonants in Tamil, but 
most scholars now assume that, despite the omission, this voicing is old and 
goes back to prehistoric times.^® 

The phonemic substitutions in sandhi and word formation are — ^in accord- 
ance with the Pratiiakhya-s and some grammars — seen as ‘change’ or ‘becom- 
ing’ (I 189 dkum, 259 akutd). Occasionally, the arbitrary procedure calls to 
mind the cynical remark ascribed to Voltaire in which he refers to etymology 
as a study “where consonants count little and vowels nothing:” tonnuru ‘90’ 
is obtained jfrom onpan "9'+paktu ‘10’ and tolldyiram ‘900’ from onpdn ‘9’-f- 
nuru ‘100’ by ad hoc sound for sound substitutions, although tonnuru obviously 
contains nuru 100’ and tolldyiram dyiram ‘1000’ (I 445+463). Another pho- 
nemic tour de force instead of proper analysis is the explanation of Cdttantai 
‘Cattan’s father’ from Cdttan + tantai ‘[his] father’ (Cdttlan] + {t]antai) where 
the occurrence of entai ‘my father,’ nuntai ‘your father,’ etc. should have 
suggested a bound form tai ‘father’ (1 348). 

The number of cases (seven, or eight if the vocative is counted), it seems, was 
established under the mfluence of Sanskrit. Instead of one Tamil instrumental 

J. F tllt ozat, JA 229.516, fn. 1, sees the source of the division in mey ‘body, 
consonant’ and uyir ‘life, vowel’ in Aitai‘eya Aranyaka II 2, 4 where body, self and 
breath are equated with consonant, voice and sibilants ; but see H. Scharfe in 
German scholars on India (Benares, 1973), p. 272. 

A. CHAimBA Sekbar, IL 16.302—305, and H. Schabfb in German Scholar, 
vol. I, p. 270-273. Hence tu can be called a ‘one-letter word.’ 

T.P.MEENAKSHiSTOmABAM, A history, p. 73; F.B. J. Ktjtper, HJ 2.191-207 
and K. Zvelebil, Comparative Dravidian phonology (The Hague, 1970), p. 161. 
C.R. SANEABAjir, Phonemics of Old Tamil (Poona, 1951), attempts an explanation 
on the basis of the Alpha-phoneme theory. 

K. ZvEiEBiL, Comparative, p. 79-84, with literature. 

M.B. EiviENEAxr, Lg. 29.339-341. 

Grammars of the Dravidian languages 


case one might postulate an instrumental ending in -dnldl and a sociative ending 
in -otuldtu, while the genitive in -atu may not be a case at aH but a neuter 
adjectival form. The cases are named after their most common suffix (the end- 
ingless nominative is called eluvdy ‘beginning, source’) and are assigned stand- 
ard values.2o Basic notions like agent, instrument, etc. are mentioned (II 108; 
H 80 has even karwmam ‘object,’ Sanskrit harman) but are not the basis of the 
case syntax. Frequently, one case suffix is said to stand for another and the 
Tolkappiyam contains the begnmings of the theory that the case suffixes serve 
different cases and assume different meanings while they do so {verr^mai- 
rmyakkam ‘confusion of case signs’).2i All nouns belong to one of two classes: 
uyartimi ‘higher class’ (for humans) and ahtiTmi ‘non-class’ ; combining these 
classes with the grammatical number we get five pdl ‘groups:’ male sing., 
female sing., human plural, one object, objects. The verb must usually cor- 
respond to its coordinated nominative, but in exceptional cases a higher class 
noun is combined with a non-class verb (II 57). Verb forms are of two kinds: 
vimi ‘action’ proper (later called terinilai-vinai ‘explicit action’) and hurippu 
‘implied action’ or ‘nom pronominalisd’ which is derived from nouns and has 
no tense marker.^^ Many verb forms are taught through prototypes using the 
root cey ‘do’ as a model, inspired no doubt by the use of ^kr in early Sanskrit 
grammar (above, p. 127) ; the same technique is also used for noun formation.^s 

The classification of nominal compounds shows the influence of Mahabha^ya 
I 378,24-379,3 when Tolkappiyam II 413 lists those with the stress on the 
first member, second member, both or neither, corresponding to avyayibhava, 
tatpurusa, dvandva and bahuvrihi.^^ More frequently there are six types: 
reflecting a case relation, reflecting a comparison, verbal compounds, qualifica- 
tions, additive compounds, and possessive compounds. 

Tolkappiyan does not recognize adjectives as a separate category and this 
has been used to support the contention that Old Tamil lacked adjectives alto- 
gether.25 Whatever the validity of this claim, Sanskrit grammar has no separate 
category for adjectives either and the four word classes of Tolkappiyan (noun, 
verb, formative element, particle) are clearly inspired by the four classes found 
in Nirukta 1 1, etc. 

20 itaimt-ituv-ena as value of the genitive (II 76) reminds one of Panini V 1 16 
tad asya . . . 

21 Explained by T.P. Meekajcshisitndabak, A history, p. 101, as late specializa- 
tions of case signs ; he overlooks the same phenomenon in many other languages. 
K. ZvELEBiL, IJDL 1 (I), p. llOf., denies the equivalence of avanukJcu peru enm 
“What is the name for him?” and avd p^ru enna “What’s his name?” avanukJcu here 
is not a genitive. 

22 On the adverbial participle cf. S. Agesthlaxingom, IL 29. 1-15. 

22 H. ScHABFE in German scholars, p. 277 f. 

2^ S. VArsTAPXJBi Ptt. in C. Eunhan Baja Presentation Volume (Madras, 1946), 
p. 136-138. 

25 M. Aotkonow, IJEL 1 (11), p. 1-9. 


H. Scharfe • Grammatical Literature 

From the time immediately following the Tolkappiyam we have only 
fragments of other grammars, notably Avinayam (pre-9th cent. A.D.) which 
has influenced the author of Namiul.^® 

In the 11th century the Buddhist petty king Puttamittiran wrote a grammar 
named Viracoliya-kkarikai in honour of his overlord Viraracentira Golan; it 
was commented upon very soon by Peruntevanar, possibly a disciple of the 
author. The five sections of a proper compendium (sound/letter, word, subject 
matter, metrics, poetics^^) are here independent chapters with 181 verses in 
kattalai-hkalUturai metre; the first two sections, the grammar proper, have 83 
verses. One can put Puttamittiran’s contribution under four headings: intro- 
duction of the terminology and theory of Sanskrit grammar, frequent references 
to the Sanskrit loan words in Tamil, observation of linguistic innovations and 
a greater conciseness of rules. 

He introduced the sis hdraha-s known from Panini’s grammar's (verse 29) to 
explain the use of the eight cases. The other case suffixes are added (except in 
the vocative) to the endmgless nominative. The opposition of the unmarked 
singular versus the marked plural is explained with a ‘zero suffix’ cu (30; 
probably short for culi or cunrmm ‘zero’). The same ‘zero suffix’ is added in 
some present tense forms (66) and root nouns (62). The shifting relations of 
karaka-a and case forms are defined with the help of 23 upakdraka-a ‘accessories’ 
(38-43): ‘agent’ has five, ‘limit’ two, ‘instrument’ two, ‘recipient’ three, ‘object’ 
seven and ‘location’ four.^^ The vocative goes with the 2nd person of the verb 
only, whereas all other cases can go with any person. Tdtu (Sanskrit dhdtu) is 
the verbal stem as seen in the 2nd sing, imperative rather than the root; the 
similarity with pre-Paninian thought is accidental. From stems, e.g. un ‘eat,’ 
we can form a causative (kdritam) uttu ‘feed,’ a second causative (kdrita-kkdri- 
tarn) uttuvi ‘cause to feed’ and a third causative (kdrita-kkdrita-kkdritam) 
uttuvippi. The Tanul infinite is called tum-anta in Sanskrit terminology. The 
implied verb is, if at all, mentioned only in passing (81). The sandhi section 
notes the forms of negation (a-, an~, na-) and guna/vrddhi in Sanskrit words as 
they have become part of the Tamil language (10-12). The treatment of com- 
pounds is influenced by the Vararuci-karika-s.^o The treatment of tattita 
(taddhita) suffixes, true to its synchronistic character, covers both original 
Tamil and Sanskrit suffixes: -an in valailyjan ‘fisherman’ and -eya in Vainateya 

26 G. ViJAYAVENTTGOPAu, A modem evaluation, p. 6-26; the fragments are 
collected in M. C. Venkatacami, Maraintu pona tamil nulkal, 2nd ed. (Madras, 
1967), p. 243-256. 

27 alankaram ; there are references to Dandin’s work. 

28 avati (35) (Sanskrit avadhi) shows the influence of the Buddhist Candragomin 
(C n 1 81) ; the statement about the nominative denoting porutmdttiram (33) ‘merely 
the subject matter’ reminds one of Candragomin’s II 1 93 artha-mdtre prathamd. 

29 Of. upakdra in Vakyapadiya III 7, i3f.; 149, 

20 P.S. SUBRAHMANYA SaSTRI, JOB 2. 105-110. 

Grammars of the Dravidian languages 


‘son of Vinata’ (52-54) ; the historical distinction educated Tamil Brahmins 
would make does not exist for a naive Tamil speaker 

Puttamittiran has observed phonetic change in local dialects (merger of I 
and I, of r and c, intervocalic l>y, intervocalic y>c) and the emergence of the 
present tense suffix -him-. The palatalization of the initial /n/ if the preceding 
word ended in ili/ai (15) is very special — such forms are actually found in the 
hymns of Appa Tevaram. Puttamittiran’s rules are short and often omit the 
contexts that condition the application of rules. 

At the beginning of the 13th century A. B., the Jain Kunavira-pantitar wrote 
his Neminatam (named after the Jain tirthamkara enshrined at Mylapore) as 
an easy grammar in 95 stanzas in the venpd meter. It has only two sections: 
sound/letter and word. In many ways it returns to the concepts of the Tol- 
kappiyam (the hdraha theory is abandoned) but it retains many of the Sanskrit 
terms and deals briefly with Sanskrit sandhi and ablaut (lOf ). Modem forms 
are recognized along with the old forms ( per : peyar, potu : polutu, -otu : -otu [60] J . 

This grammar was overshadowed by the Na^nul ‘Good Treatise’ of another 
Jain from northwestern Tamilland, Pavananti, who wrote at about the same 
time under the patronage of the Ganga king Ciyakahkan. It has three sections 
(preface, sound/letter, word) with 462 verses in nurpd meter. The propaedeutic 
preface, with its 65 verses, is, in the opinion of U. V. Swaminatha Aiyar,32 a 
later addition. The oldest commentator is Mayilainatar (14th cent.). 

In his treatment of the linguistic development, Pavananti is more conserva- 
tive than Puttamittiran and Kunavirapantitar. While he recognizes the popular 
palatalizations (e. g. aintu > ancu ‘5’), he frequently returns to the Tolkappiyam. 
The same generally holds for his terminology, although he does speak of pahuti 
‘base’ (Sanskrit prahrti) and vihuti ‘modification’ (Sanskrit vihrti) and once 
even uses accu ‘vowel’ (Paninian a^). Pavananti uses the previous literature 
eclectically and makes his contribution with a lucid compilation. Among his 
well coined terms is cdrp-eluttu ‘dependent sound’ for allophones. 

Cuppiramaniya-titcitar (17th cent.) went beyond Puttamittiran in his appli- 
cation of Sanskrit terms and theories in his Piraydka-vivekam, and towards 
the end of the 17th century Vaittiyanata Navalar wrote his Ilakkana-vilakkam 
based on the Tolkappiyam, the Nannul and their commentaries. In the 18th 
century, the Italian Jesuit C. J. Beschi (who also wrote two grammars on liter- 
ary and colloquial Tamil in Latin) composed his Tonnul-vilakkam in the 
traditional style based on the Nannul. He has also left a lasting imprint on the 
way Tamil is written today by his invention of modified letters for /e/ and /6/ 
to contrast them with /e/ and /o/. 

The anonymous Lilatilakam from Kerala is a manual of a peculiar literary 
style called mani-pravdlam ‘jewel and coral,’ consisting of 151 sutra-s in San- 

Cf. the treatment of Arabic elements in Persian by Kf^nadasa (below, p. 197). 
Quoted by G. VijayaventtgopaIj, A modern evaluation, p. 4. 


H. Scharfe • Grammatical Literature 

skrit and a vrtti in Malayalam, probably by the same author ;33 historical 
references in the vriri point to the years A.D. 1375-1400. There can be little 
doubt that the term mani-pravdlam originally alluded to a kind of traditional 
South Indian jewelry encrusted with pearls and corals. The Lilatilakam reinter- 
prets the term as referring to a combination of ‘(red) rubies and (red) corals’ to 
reflect his different notion of the mani’pravdlam style as it had developed in 
Kerala: not a contrastive use of Sanskrit and Dravidian but an amalgamation.^^ 
The style has been called macaronic, but perhaps this is not correct. Its charac- 
teristic is not a supplementation of an author’s Sanskrit vocabulary with verna- 
cular sterns^ or a Sanskritized vernacular — ^it is a real ‘hybrid’ style with two 
separate grammatical systems; Sanskrit words (coral) and Tamil words (ruby), 36 
with their respective endings, are blended in a sentence, with a preponderance 
of the vernacular (bhdsd).^’^ The words of the hJiasd are either regional, 38 derived 
from Sanskrit or identical with Sanskrit. 

In its phonology Lilatilakam considers the sounds that are peculiar to Tamil 
and cannot be derived from either Sanskrit or other vernaculars (Kannada, 
etc.). The phonemic distinction of r and r {uri ‘measure:’ uri ‘hoop’) is recog- 
nized; similarly the distinction of n and n is phonemic in Tamil but not in 
Sanskrit (p. 79 - 82).38 Phonemes typical for Sanskrit (aspirate stops, s, s, etc.) 
have entered the hhdsd through the Sanskrit words introduced by the members 
of the three upper classes (p. 115f.). 

33 P.N. Elamkctlam PrLLAi in the preface, p. 14, to his edition of the Lilatilakam 
(Kottayam, 1968). It is little noticed that in North India there was a similar mixed 
style that merged Sanskrit and a Prakrit dialect. Bhoja, in his literary manual 
Sarasvati-kanthabharana 11 27, calls it samhirna [jdti] and compares it to a mixture of 
sesame and rice grains. Old Javanese poetry, too, mixes the vernacular (Javanese) 
with Sanskrit (K.M. Panikkab in C. Kunhan Raja Presentation Volume, p. 65-69). 

34 In Viracoliya-kkarikai 180 and 142 comm, mani-ppiravdlam denotes instead a 
Sanskritized Tamil diction. 

36 Instances of macaronic style (which was permitted in certain genres) are 
kelanti ‘crying’ from keluha ‘cry’ with a Sanskrit participle suffix -anti and pupukire 
‘they entered’ from pukuka ‘enter’ with a ‘Sanskrit reduplicated perfect’ form 
(p. 75f. of P.N. Euamktjlam Pillai’s ed.). Language switching and macaronic 
forms are presently characteristic of the conversational style among educated South 
Indians — only that English has taken the place of Sanskrit. 

38 The author regards his language (Old Malayalam; Kerala-bhasd) as a Dravida 
language and hence also calls it Tamil (Tamil is an apahhramsa of Dravida); he 
uses Dravida in a narrow sense, comprising only the language of the three kingdoms 
(Cola, Pantiya and Kerala) and excluding the languages of the Karnnatakar and 
Telunkar which other authors of his time would include (p. 34; cf. A. Chlanuba 
Sekhab, XXth AIOC vol. H, p. 261—266 and K.M. Geobge, Studies in Indian 
Linguistics [Fs. M.B. Emeneau, Poona/Annamalai, 1968], p. 95-98). 

37 Laatnakam,p. 32; 37; 57-63 (= sutra-s 3-10). 

38 These again can be ‘pure,’ derived from another hhdsd e.g. Kannada) or identi- 
cal with such a hhdsd (p. 70f. ; cf. K. KiJNJTOrNi Raja, IL 30. 70-72). 

33 C.R.Sai^babait andK.M.N.MENON, BhV 20/21.392-394; L. V. Rajmaswami 
Aiyab, IL 25. 270-274. 

Grammars of the Dravidian languages 


The presentation of case suffixes follows the pattern found in the Tolkappi- 
yam and the Nannul with mi n or adjustments (genitive nnu and locative iljvil 
instead of atu and Jean), Tamil cases should be used according to Tamil syntax, 
Sanskrit cases according to Sanskrit syntax; the opposite is condemned: the 
instr. in Jcdnta-noJclcindt-utsuJcaTii, analogous to Sanskrit Jcdnta~daTsa 7 i£,ii 6 t 8 uJc€L 7 ii 
‘eager to see the beloved/ is a Sanskritism just like the acc. in divasatte ninmn 
‘he stood a day long’ (p. 87). Surprisingly, Lfiatilakam denies that the -e in 
avane-Jeontu ‘with him/ lit. ‘having taken him/ is an acc. suffix; it assumes 
instead a ‘union vowel’ (samdhdyaJca) -e- (p. 88). 

A.R. Rajarajavarman (1863—1918)^^ combines in his Kerala-paniniyani^^ 
(written in modem Malayalam) Panini-like description of his mother tongue 
with an historical insight into the Dravidian nature of the language. The in- 
fluence of R. Caldwell’s Comparative Grammar of the Dravidian Languages 
and H. Gundert’s Malayala-vyakarana is evident, but he avoids simple adop- 
tion of Western terminology as much as he avoids a mechanical application of 
Paninian categories. Rajarajavarman recognizes that the Malayalam cases do 
not correspond one by one to the Sanskrit cases. He separates the ‘agency’ case 
in -dl from the ‘society’ case in -otu and correctly analyzes ‘quasi-case forma- 
tions’ hke atil-ninnu ‘from it’ as a loc. atil plus ninnu resulting in a pseudo- 
ablative. This grammar has influenced the course of modem Malayalam litera- 
ture, steering a middle course between Sanskritizing and colloquial tendencies 
evident in the late 19th century writing. 

The oldest Telugu^^ grammar would be the Andhra-^abda-cintamam (82 to 90 
arya verses = 274 sutra-s in Sanskrit) if it is correctly ascribed to the poet 
Nannaya[bhattu] (11th cent. A.D.); but that is disputed. It was commented 
on in Telugu prose by Elakuci Balasarasvati (c. 1550-1600), Appakavi, (c. 
1600-1670) and Ahobalapati (c. 1700). Mulaghatika Ketana (1220-1300) wrote 
his Andhra-bhasa-bhusana in 192 Telugu verses, claiming this to be the first 
Telugu grammar. A few decades later, Atharvanacarya composed (besides a 
Telugu grammar in Sanskrit verses called the Vikrti-viveka supplementing the 
Andhra-^abda-cintamani) his Trilinga-sabdanu^asana, an essay on the origins 
of the Telugu language. In the 19th century, the Bala-vyakaranamu by Para- 
vastu Cinnayasuri, in Telugu sutra-s and arranged topically like the Siddhanta- 
kaumudi, was so influential that its standards were even applied retroactively to 

The author follows Kumarila (Tantravarttika on 1 3, 9) in ridiculing such forced 
Sanskrit-Tamil etymologies as cor ‘rice’ from cora ‘robber’ or vayaru ‘stomach’ 
from vairin ‘enemy’ (p. 72f.). 

41 Biographical details in K. Kunjunni Raja, The contribution of Kerala to 
Sanskrit literature (Madras, 1958), p. 256f. 

42 The first edition appeared in 1871 ; the 2nd edition of 1892 replaces the sutra-s 
with 194 stanzas. An extensive commentary by the author explains and illustrates 
the rules. 

43 Sarveswaba Shabivia Pebi, ZDMG Supplement H, XVIII. Deutscher Orien- 
talistentag, p. 384-389. 


H. Scharfe • Grammatical Literature 

editions of Old Telugn poetry. B. Sit^amacarynlu’s Prau^a-vyakaranamu, 
also called Tiilinga-lak^na-sesamu, supplements the Bala-vyakaranamu and 
offers a wealth of illustrations. It was published in 1885. 

Nagavarma (c. A.D. 1150) wrote the two oldest grammatical treatises^ on 
Kannada (Canarese): the ^abda-smrti (in Old Kannada, 96 verses) is part of 
his literary manual Kavyavalokana, whereas his Karnataka-bhasa-bhusana (280 
sutra-s with a vrtti, both in Sanskrit) is an independent work; one of its ten 
sections deals with the kdraJca concept of semantics. From the 13th century we 
have Kesiraja’s ^abda-mani-darpana in Old Kannada (322 sutra-s in kanda 
metre and a vrtti) ; this comprehensive grammar of the Kannada language 
forms the basis for F. KitteFs Grammar of the Kannada Language.^s Xhe 
Karnataka-sabdanu^sana by Bhattakalanka Deva is dated A.D 1604. Its 
concise 592 sutra-s, the vrtti and vydJcJiyd are all written in Sanskrit and the 
author frequently quotes the southern version of the Jainendra grammar.^® 
Krsnamacarya from Srirahgapatiana wrote his Hosagannada nu^ganna^, also 
titled Grammar of the Modem Canarese Language, at the urging of a British 
officer. After some years’ delay, in 1838, the grammar was printed in Madras 
where the author lived for many years as a pleader at the courts. Its special 
merit is the study of the relation of Kannada to Sanskrit and Tamil, which 
shows the influence of F. W. Ellis. Phonetic correspondences between Tamil and 
Kannada words allow him to derive Kannada words from Tamil. A weakness is 
the occasional neglect of Old Kannada which lets him project developments 
within Kannada into the prehistoric period. The division of the vocabulary 
into five classes: tatsama ‘identical [with Sanskrit],’ tadbhava ‘derived [from 
Sanskrit],’ dUya ‘vernacular,’ anyadesya ‘from another vernacular’ and grdmya 
‘vulgar’ shows, perhaps, the influence of Ketana’s Telugn grammar.^® 

44 Nipatunga’s Kavi-raja-marga (9th cent. A.D.), though it contains some gram- 
matical observations, is primarily a manual of poetics. 

45 F. Kittel, a Grammar of the Kannada Language (Mangalore, 1903). On page 
3 of his grammar, Kittel summarizes the history of grammatical literature in 

46 K.B.Pathak, ABORI 13.27. 

47 T.K. Skeekantaiya, Studies in Indian Linguistics [Fs. M.B. Emeneau] 
(Poona/Annamalai, 1968), p. 322-331. 

48 Cf. above fh. 43. 

Chapteb XIV 


A strong case can be made for the importance of princely patronage of gram- 
matical studies. We can see three spurts of activity: in the 5th century A.B. 
(Candragomin, Bhartrhari, Devanandin), the 11th to the 13th century A.B. 
(Kaiyata, Bhoja, Hemacandra, Kramadisvara, Anuhhutisvarupa, Vopadeva, 
Purusottama, Trivikrama, Damodara) and in the 17th century A.B. (Bhattoji 
Biksita and his school, Markandeya, Mirza lOian), which coincide with the 
Gupta dynasty, the prosperity of the Hindu kingdoms before the MusHm 
conquest and the height of the Mughal rule. Both Sanskrit and Prakrit studies 
profited from the favourable conditions. The political fragmentation of India 
before the Muslim conquest may well explain the creation of several original 
(i.e. non-Paninian) systems of Sanskrit grammar. The renaissance of Paninian 
studies led by Bhattoji Biksita (17th cent. A.B.) and his school coincided with 
the rise of Mahratta power which eclipsed the traditional patrons of several 
grammars; this renaissance swept these schools from the main part of India 
into residual pockets in Bengal, Kashmir, etc. 

The first of these grammars is the Sarasvati-kanthabharana written by 
Bhoja, the Emg of Bhara in western Madhya Pradesh.^ As the author says in 
his Rajamrgahka, the grammar was composed in A.B. 1042. In eight books of 
four chapters each (with more than 6000 rules altogether), he incorporates the 
content of Katyayana’s varttika-s, the unadi-sutra-s, the word lists, the Bhatu- 
patha and the metarules. The arrangement is topical, beginning with definitions 
and metarules and ending with Vedic and accent rules. The other main topics 
are: primary suffixes, nominal and verbal endings (in the same bookl), composi- 
tion, feminine formation, secondary suffixes, and sandhi. Occasionally Bhoja 
recognizes forms or meanings that were condemned by earlier authorities but 
had nevertheless gained acceptance from Sanskrit writers. To give only one 
example: while Mahabhasya II 399, 21 f. teaches in the meaning ‘unable 

to bear snow,’ Sarasvatikanthabharana V 2 226 has for it the meaning ‘able to 
bear snow.’ A commentary on this grammar, called Hrdaya-harim, was written 
by Narayanabhatta, possibly a courtier of the king. 

The grammar Sarnksipta-sara by Eframadi^vara (12th cent. A.B. or earlier 

^ King Bhoja also wrote, under the same title, a manual of poetics. 

^ Th. Zachabiae, BB 6.22-63. Kramadisvara was probably a native of Bengal 
PCi. KTitti-Bolci, Les graromairiens prakrits (Paris, 1938), p. 130 and N.N. Bas- 
GUPTA IC 5,358]. 


H. Seharfe • Grammatical Literature 

consists of c. 4000 siitra-s in eight books. Jumaranandin (13th cent. A.D.?) 
added a commentary, Rasavati (or is it only the revision of an existing com- 
mentary?). and the school is often called Jaumara after him.^ The arrangement 
is very original: after the sandhi rules the verb is dealt with first ; then the nouns 
are built up from primary and secondary noun formations, through kdraka 
relations, to inflected case forms and finally compounds. The eighth book is set 
aside for the Prakrits (cf. Hemacandral). 

An imconventional grammar is the XJkti-vyakti-prakarana of Pan^ta Gamo- 
dara who lived in Benares during the first half of the 12th century A.D. and 
tutored the sons of the Gahadwala king Govindacandra. Its 50 karika-s in dryd 
meter deal in five chapters with verbs, the use of cases, the combination of 
semantic functions, letter writing in general and the writing of business letters 
in particular. About half of the appendix on transitive and intransitive verbs 
and the last two chapters of the author’s own commentary are lost in the only 
existing manuscript. Damodara frequently refers to the Katantra. 

Damodara’s idea is that the colloquial language (uUi) of his time, with all 
its deplorable deviation from Sanskrit, is but Sanskrit in disguise and can 
acquire dignity as the base for a renewed use of Sanskrit. This is possible be- 
cause the noun and verb forms of Sanskrit have their correspondences in the 
colloquial and do not differ in meaning but only phonetically {ydny eva sarnskrta- 
hhdsdydm sup-tin-antdni paddni tdny evdpahhramse ’pi, na cdrthe mandg api 
bhedah, kevcdam aksaresu viparyayah; stanzas 6/7 commentary). A problem 
arises when sometimes the colloquial, being simpler and more analytical than 
Sanskrit, offers only insufficient clues as to the correct Sanskrit form: in the 
distinction of genders, in case morphology and regarding the several past tenses 
of Sanskrit. The author’s use of coUoquial forms (e.g. karaii ‘he shaU do’ > im- 
perative/optative) as the point of departure for his transfer rules gives us 
virtually a grammar of the 12th century language of Benares, i. e. Old Kosali 
or Eastern Hindi, a forerunner of Tulsidas’ Awadhi by four hundred years.^ 
This testimony no doubt constitutes for us the book’s greatest value. The two 
chapters on letter writing stand in an old tradition as the sasanddhikdra of the 
Arthasastra II 10 shows.® 

Vopadeva (late 13th cent. A.D.), a member of a family of physicians in the 
service of the Yadava kings in Maharashtra, is the author of the Mugdhabodha 

3 The subcommentary by Goyicandra (on the books I to VII only) is renowned for 
its lucid treatment of syntax in its kdraka section. 

^ The Mugdhavabodhamauktika (written in A.D. 1394) similarly supplies us 
indirectly with a sketch of Old Gujerati grammar (G.A. Gexebson, JRAS 1902. 
537-555); cf. Sadhusundara Gani’s Uktiratnakara (ed. Jinavijaya Muni; Jaipur, 
1957) written in the time of Akbar with material from Rajasthani and neighbouring 

5 Cf. the Lekha-paddhati, ed. C.D. Daiax and G.K. Shbigondeeab. (Baroda, 
1925), and the remarks of O. Stein, ZII 6.45-71 and H. Sckabfe, Untersuchungen 
zur Staatsrechtslehre des Kautalya (Wiesbaden, 1968), p. 68-75. 

other systems of Sanskrit grammar 


(1184 sutra-s in 26 sections). Besides this grammar, a root dictionary, etc., he 
has to his credit works on medicine, dharmasastra and literature. The author’s 
religious fervour is evident in the examples of his grammar that use the sacred 
names of Visnu and Siva whenever possible.^ Vopadeva’s most striking innova- 
tion is his algebraic terminology that goes far beyond that of Panini and Beva- 
nandin; even the sound table has been altered to fit the peculiar design of the 
grammar. The author’s originality, in the long run, was detrimental to his 
success: literary commentaries written with references to Panini’s grammar and 
the grammatical classics themselves were closed books to the students of the 
Mugdhabodha. The Mugdhabodha aims at brevity and simplicity; inflection 
rules are illustrated with partial paradigms, unusual forms and the Vedic 
language including the accents are all but neglected.? The arrangement of topics 
is derived from the Katantra and Dharmakirti. As Vopadeva could formulate 
his rules to suit his arrangement (giving the hdraha relations their duel), one 
can imagine that the Mugdhabodha only narrowly lost out to the Siddhanta-kau- 
mudi and the Paniniya-s. 

The origin of the Sarasvata-vyakarana is shrouded in mystery; for the 700 
Sarasvatl-sutra-s^ which, according to one tradition, were revealed by the 
goddess Sarasvati to Anubhutisvarupacarya (13tli-14th cent. A.D.) were, 
according to another tradition, but a later creation.^ The leading text of the 
school is Anubhutisvarupacarya’s Sarasvata-pralrriya (1494 sutra-s). In the 
preamble the author claims: Sdrasvatim rjum hurve 'prahriydm ndtivistardm 
“I straighten out the procedure that goes back to Sarasvati without undue 
prohxity.” This statement is incompatible with the tradition of divine revelation 
to Anubhutisvarupacarya. Since Anubhutisvarupacarya does not refer to the 700 
Sarasvati-sutra-s (nothing is said about doubling their number!), his words can 
be taken in either of two ways: sarasvati may be used poetically as a synonym 
for ‘speech,’ meaning that the author simplified the complex structure of the 
language ; or it could refer in an abbreviated form to an older grammar, namely 
the Sarasvati-kanthabharana (though at first glance no special relationship be- 
tween these two works is visible). In his arrangement Anubhutisvarupacarya 
follows Dharmakirti and Vopadeva very closely but stays within the conven- 
tional terminology. The Sarasvata grammar enjoyed for some time great 
popularity and was patronized by both Hindu and Muslim princes. It is striking 

^ Vopadeva was outdone in this respect by later sectarian grammarians, e.g. the 
authors of the two Harinamamrta-s, Rupagosvamin and Jivagosvamiti (15th and 
16th cent. A.D.) who use God’s names even for technical expressions (e.g. Furasotta- 
ma — ‘long’) as the mere uttering of the names was regarded as beneficial. 

As happened with the Katantra, later followers supplemented the work with 
further rules to cover every odd form. 

® These were commented on by Ramacandra^rama (not later than the 16th cent. 
A.B,) in his Siddhanta-candrika. 

^ Harapbasaba ShAstbi, a descriptive catalogue (Calcutta, 1931), p. 136. 

For yet another explanation see Yxjdhisthib MImamsax, Samskrt vyakaran 
kaitihas, I, p. 627. 


H. Scharfe * Grammatical Literature 

that all earlier authors of the school were samnydsin-^, religious devotees who 
renounced the world ; it would seem that a simplified grammar met the samnyd- 
siTZ-’s need of a working knowledge of Sanskrit just as it satisfied the educational 
desires of the ruling class. The great number of textbooks in the form of sub- 
commentaries written by court pandits shows that the pressure to ‘publish or 
perish’ existed long before our time. 

In search of prestige the rulers of Mithila and of Gooch Bihar commissioned the 
Saupadma grammar by Padmanabhadatta (14th cent. A. D.) and the Prayoga- 
ratna-mala by Purusottama (16th cent. A.D.). Efforts to teach Sanskrit by the 
‘direct method’ produced conversation manuals: Varadaraja’s Girvana-pada- 
manjari andKasmatha’sPradipa;ii but we must also consider that traditional 
pandits probably taught Sanskrit by the direct method since time immemorial, 
even without specially devised conversational manuals. 

Habapbasada Shastbi, a descriptive catalogue, p. cxvii f. 

Chapter XV 


For the educated Brahmin^ the ‘common’ (prdlcHa) language was meaningful 
only insofar as it reminded him of the correct Sanskrit form familiar to him; 
but common speakers would understand the Prakrit words immediately and 
would even assert that for them the Sanskrit forms conveyed meaning only 
through the Prakrit forms they brought to their mind (Vakyapadiya 1 151-155). 
The poet Vakpati (8th cent. A.D.) thus regards Prakrit as the source of all 
languages, including Sanskrit: 

aayaldo imam vdyd visanti etto ya nenti vdydo 
enti samvddam ciya nenti sdyardocciya jaldim 

Gaiidavaho 93 

“AU languages enter this [Prakrit] and ail languages take their start from this: 
the waters enter nowhere but into the sea, and start from nowhere else than from 
the sea.” 

Conversely, Hemacandra in Ms commentary on his own rule VIII 1 1 explains 
that “Sanskrit is the base ; what originates in it or comes from it is base-derived” 
(jprahHih sarp^hrtam; tatra hhavam tata dgatam vd 'prdkrtam). 

It is hard to believe that the early J ains and Buddhists, who pioneered the 
translation of canonical texts from one vernacular to another, should not have 
given any thought to grammar; but other than occasional remarks in their 
scriptures they have left us no systematical treatises. 

The oldest pieces of Prakrit grammar are perhaps the fragments preserved 
in chapter X"^I of the Bharata-Xatya^astra (hardly later than the first cen- 
turies A.D.). The first fragment, the stanzas 6 to 9 composed in the popular 
dryd meter, gives phonemic rules for the conversion of Sanskrit words into 
Prakrit; the most interesting featiue of these stanzas is that they are written 
in Prakrit themselves.^ In the Sanskrit stanzas that foUow (10 to 23), the 
examples are displayed more prominently than the rules; the rules are abstract- 
ed from the examples. Evidently the Natyasastra, being a manual for actors, 
was less interested in comprehensive granunatical rules than in a number of 
characteristic expressions. In the stanzas 25 to 56, Sanskrit and the various 
vernaculars are assigned to stage personalities: gods and bra hm i n s speak 

i Such a tradition of grammatical rules in Prakrit may have existed among the 
Jains (A. X. Upadhye, ABORI 13. 45 f.) ; a few such rules are quoted in later texts. 


H. Scharfe • Grammatical Literature 

Sanskrit, employees of the royal harem Magadhi, and guild masters Ardhama- 
gadhi, etc. On a different level again, stanzas 57 to 61 advise the actor to play 
persons from Bihar with an abundance of /e/ sounds, people from Sindh with 
many /u/ sounds in their speech, etc. : a superficial mimicking of the real-dialects. 

A much more detailed account of Prakrit is the Prakrta-prakasa (or Prakrta- 
laksana-sutra; title uncertain) attributed to the elusive author Vararuci; the 
eight books (with altogether about 420 sutra-s in Sanskrit) in fact deal only 
with Maharastri. The great number of permitted duplicates and the striking 
correspondences with forms occurring in Hala’s collection, Sattasai (2nd cent. 
A. D,), suggest that the rules were abstracted &om a similar collection of popular 
songs with regional grammatical differences. 

In the 7th century A.D., the Prakrta-prakasa was commented on by the 
rhetorician Bhamaha whose text however includes two additional books; one 
on the Paisaci dialect and the other on Magadhi, both of which are unknowm to 
the other and much later commentators of the text.^ Some time after Bhamaha, 
book V was split into two when yet another book (on Sauraseni) was added, 
making a round number of 12 books. 

The Prakrta-prak^a evidently presumes a knowledge of Paninf s grammar, 
and lists, without any introduction, transfer rules that allow the connoisseur 
of Sanskrit to form correct Maharastri poetry. The starting point is the Sanskrit 
language in its pre-use stage: the suffix of the nom. sing, is still that of the 
gen. sing, still when they are replaced by d and ssa. Instead of dative 
suffixes the genitive suffixes are used and plural forms replace dual forms. The 
difficulty of formulating precise transfer rules leads to an excessive use of 
'often’ and 'or’ in all Prakrit grammars; the practice has its forerunner in 
Panini’s treatment of Vedic forms (6aMam chandasi Panini II 4 39, etc.). 

An 'eastern school’ of Prakrit grammarians expanded Vararuci’s opus closely 
following, in the main part, Vararuci’s rules for Maharastri and then dealing 
with the stage Prakrits similarly to the Natyasastra; they add a treatment of 
Paisaci and Apabhramsa. The oldest of the grammars preserved^ is Purusot- 
tama’s^ Prakrtanusasana (12th cent. A.D.) preserved in a single manuscript; 
Markandeya wrote his Prakrta-sarvasva in the 17th century or earlier, remark- 
able for his philological acumen and reliability; Rama^arman’s Prakrta-kalpa- 
taru (17th cent.) again survives in a single manuscript. These authors lived at a 
time in which direct observation of spoken Prakrits can be ruled out; they had 
to rely mstead on the grammatical tradition (which they often misunderstood) 
and on a study of available manuscripts of the Prakrit classics (with a broad 

2 Vasantaraja’s Prakria-samjivani (14th— 15th cent.), Sadananda’s Prakrta-subo- 
dhini, anonymous Prakrta-manjari, Narayana Vidyavinoda’s Prakrta-pada and 
Ramapamvada’s Vriti ( 1 8th cent. ) . ’ 

^ The frequent references to the teachings of ^akalya suggest that an author of 
this name wrote a Prakrit grammar that is now lost (L. Nitti-Dolci, Les gram- 
mairiens prakrits, p. 95). 

^ He is also the author of a work on Panini’s grammar (above, p. 174). 

Grammars of the Middle Indo -Aryan dialects 


spectrum of variant readings). The problem of how far we should go in correct- 
ing the Prakrit literature to conform with the rules of the grammarians is not 
easy to decide; the manuscript fragments of some dramas found in the sands 
of Turkestan are actually earlier than any of these grammarians. 

The Jain Hemacandra Suri (A.D. 1089-1172) taught the Prakrits through 
transfer rules as did Vararuci; it was only logical that he offered these 1119 
rules in the eighth and last book of his Sanskrit grammar (above, p. 169).5 The 
transfer rules follow the rules on Sanskrit grammar and close with the statement 
that in all remaining respects Prakrit is like Sanskrit (sesarii sarnskrtavat 
siddham). Hemacandra’s formulations depend heavily on Vararuci, with many 
additional rules on the 'basic Prakrit' (i. e. Maharastri) inserted here and there. 
He is the first author we know of to state the obvious rule that a long vowel 
before a consonant cluster is shortened in the transformation: VIII 1 84 hrasvdh 
samyoge, Hemacandra’s treatment of Magadhi, Pai^aci and Sauraseni shows the 
influence of a lost treatise of which we have a reflection in the commentary of 
the Jain Namisadhu (A.D. 1069) on Rudrata's Kavyalamkara II 12. More 
original are his contributions concerning the language of the canonical Jain 
scriptures, the Ardhamagadhi, which he calls drsa '[language] of the saints’; 
almost all the special forms have been verified from the texts. Apabhramsa too 
receives a detailed treatment, illustrated with many stanzas called dohd taken 
from then current poetry. Differing from other Prakrit grammars, Hemacan- 
dra’s Apabhramsa appears more as one well defined language, even though 
dialectal differences can be found in the illustrations; it is a forerunner of Old 

Kramadi^vara (12th cent. A.D. or earher) also treated Prakrit in the eighth 
book of his Sanskrit grammar, the Sarnksiptasara (above, p. 187 f.). But the 
eighth book has not enjoyed the same popularity as the rest of the work and 
manuscripts of it are extremely rare. Kramadi^vara based his work on Vararuci, 
whose description he tries to shorten in several ways. The optional replacement 
a>i taught by Vararuci for several words (I 3 id isat-pahva-svapna-vetasa- 
vycijana^-MrdangdngdTesu) is abbreviated to Vill 1 2 it pakvddeh jij [for the /a/] in 
pahva, etc.” (e.g. pikka as well as pakka 'ripe’). It is interesting to see that 
Hemacandra VIII 1 47 takes a middle position: he lists three nouns and avoids 
the use of 'etc.’ Kramadl^vara VIII 2 6f. joins Hemacandra VIII 1 180 in 
recogniziag the ya-sruti, i. e. a hiatus-removing /y/ inside a word (e. g.nagaram > 
naaram>nayaram). The few obvious similarities and deep differences do not 
yet allow any inference on the relative chronology of Kramadi^vara and 

The Prakrta-^abdanu^asana of the Jain Trivikrama? (13th cent. A.D.), with 

5 This book has its own two commentaries by the author; the shorter one is 


= The gap between Hemacandra and the oldest Gujerati texts is bridged by the 
material found in the Mugdhavabodhamauktika (above, p. 188, fn. ). 

7 A.N. TJpadhyb, BhV 2, 2, p. 160-176. 


H. Scharfe • Grammatical Literature 

the author’s vrtti, consists of c. 1036 sutra-s with a technical terminology 
that includes newly defined determinatives. Some scholars believe that the 
sutra-s are metrical and should be so written; but as sutra-s often straddle the 
assumed metrical divisions, this does not seem to be a good idea. Some com- 
mentatons on the text suggest that the sutra-s are not really Trivikrama’s but 
VahniM’s. But there seems to be no doubt that both sutra-s and commentary 
depend heavily on Hemacandra, and the author himself acknowledges his debt 
to his predecessors down to Hemacandra, thus excluding any intermediary. The 
third book of Trivikrama’s work is valuable because of the many Apabhramsa 
stanzas quoted in it. 

Eishikesh Sastri’s Prakrta-vyakarana, with English translation (Calcutta, 
1883), treats basically Maharastri and refers to other dialects in footnotes. Eules 
are formulated for multiple application only, whereas unique developments 
are listed in the tables. Infiection rules are supplemented by paradigms under 
the influence of European grammars. Probably also a recent work is Canda’s 
Prakita-laksana, although its editor, A.F.R. Hoemle (Calcutta, 1880), believed 
he had a text of the 3rd century B.C. ; no manuscripts of it have been found 
that are earlier than the second half of the 19th century A. D. 

It is possible that Buddhaghosa (5th cent. A.D.), the Buddhist commentator 
of Pali texts, refers to a lost Pali grammar when he discusses grammatical 
questions. But the first Pali grammar we have is the Kaccayana-vyakarana 
^vritten between the 5th and the 11th centuries A.D. in the Pali language; the 
oldest known commentary on it is Vimalabuddhi’s Nyasa (11th cent. A.D.).Its 
four Icappa-s, with c. 675 sutta-s, deal with phonology, nouns (including hdraka-s, 
secondary word formation and compounds), verbs, and primary word forma- 
tion. The author relied on the Astadhyayi as well as the Katantra.® The Sanskrit 
influence is visible in the recognition of a separate dative case even though its 
forms are identical with those of the genitive ; only rarely do original dat. sing, 
forms in -dya occur. The different case suffixes of the various noun classes are 
derived from a set of standard suffixes by substitutions. Among the several 
recasts of this grammar, the Eupasiddhi of Buddhappiya Diparnkara (late 13th 
cent. A.D.) is the most prominent; commentaries on Kaccayana’s grammar 
exist both in Pali and in Sinhalese.® 

Aggavamsa from Aiimaddana in Burma was the teacher of I^g Narapati 
Sithu of Pagan. He composed his voluminous Saddaniti in the year A.D. 1154 
and a copy of it was soon taken to Ceylon. The work was well received in both 
countries. Aggavamsa largely follows Kaccayana but achieved a much more 

® E. O. Peanke’s attempt to show influence of the Kasika (Geschichte und Kritik 
der einhehnisehen Pah-Grammatik und -Lexicographie, p. 17—19) is not convincing. 
Illustrations like pdtJiayati mdnavakam vedam, which the Ka4ika on I 4 52 copies 
from Candravrtti on H 1 44, are clearly older than the Buddhist Candragomin and 
cannot prove Feakke’s point even if they should have influenced Kaccayana’s for- 
mulation of rule n 6 30. 

® A rich bibliography on Pali grammar is given by D. L. Barua, IC 15. 194-202. 

Grammars of the Middle Indo-Aryan dialects 


complete description of PaU than the latter. The work consists of three main 
parts: the Padamala, a detailed morphology of the Vord and paradigm’ type 
beginning with the verb (of. Kramadisvara!) followed by norm, pronoim and 
numeral; the Dhatumala, a root list^° with a comprehensive survey of the 
attested verbal and nominal derivatives and compounds (cf. Maitreyaraksita’s 
Dhatupradipa!); the Suttamala, which in 1347 sutta-s covers the same ground 
as the two previous parts, but this time in the ‘item and process’ manner, 
closely following Kaccayana.^i 

Moggallana from the Thuparama monastery in Anuradhapura wrote his 
Magadha^s Saddalakkhana during the reign of Parakkamabahu I (A.D. 1153- 
1186). In six Jcanda-8 he treats phonology and metarules, noun inflection, com- 
pounds, secondary noun formation, secondary roots and primary noun forma- 
tion, and verb inflection. The influence of Candragomini^ is evident in the 
avoidance of several technical terms, e.g. the hdraha-&; the syntactic rules 11 
2-42 are strikingly similar to Candragomin’s II 1 43-98. In addition to Moggal- 
lana’s own commentaries vutti and pancihd (the latter is lost), there is a large 
body of literature on this system, both in Pali and in Sinhalese. 

Though the specialists evidently knew Sanskrit, it was less important for the 
Buddhist co m munities of the Theravada tradition than Pah. It is not surprising 
therefore that the Pali grammarians did not derive this canonical language from 
Sanskrit ; nor did they teach it in a transfer grammar based on Sanskrit. Though 
they depended totally on the known body of Pah hterature, their subsequent 
influence on the canonical texts must be considered in any linguistic-phflological 
study of the Theravada canon. 

A Sinhalese classic is Vedeha Thera’s ( ?) grammar of the Old Sinhalese poetic 
style (Elu), the Sidat-sahgarava, written in Elu in the 13th century A.B. 
Besides Panini, Katantra and MoggaUana, it is influenced by the Tamil gram- 
mar Viracoliyam and, like the latter, includes the elements of poetics. In the 
traditional Tamil way consonants are hkened to the ‘body’ and vowels to ‘life’ 
(gatakuru and panakuru; gdtrdksara and prari(iA’5amin Sanskritized Srohalese).^^ 

On an old Pffi root list cf. R. O. Fbanke, Album [Fs. H.] Kebn (Leiden, 1903), 
p. 353-356. The root list attached to Kaccayana’s grammar is a later addition. 

Aggavarnsa breaks up the long chapter on nouns, and treats kdraka-s, com- 
pounds and secondary noun formation separately. The kdraka-s (with the sole excep- 
tion of okdsa 572) receive dual values (548-555), e.g. yo kurute yo vd jdyatiso kattd 
„Who does or who is born is an agent” ; Kaccayanahas only two such formiilations 
(Kaccayana n 6 1 and 6). 

With the Buddhist tradition Moggallana believes Pali to be the language of 

13 R.O. Fbaitke, JPTS 1902-1903.70-95. 

1^ C.E. Godaxtjmbuba, BSOS 11.837f.; H. Gunthub, ZDMG 96.84r-97. 



Because of the Indian preoccupation with the classical languages and perhaps 
partly due to a lack of patronage in times of foreign domination, grammatical 
descriptions of several modem Indian languages were first undertaken by 
European missionaries and scholar-administrators. The need was greatest in 
the early phases of European influence and decreased later as the local people 
acquired a worldng knowledge of the European language spoken by their 
superiors. These works, e.g. Father Stephens’ Konkani grammar (16th cent. 

J. J. Ketelaer’s (i.e. Kettler) Hindostani grammar (1715)2 ^nd Manoel 
da Assump^am’s Bengali Grammar (1734), ^ fall outside the present survey. 

But already during the rule of Aurangzeb and, we can further specify, prior 
to the year 1676, Mirza Khan Ibn Fathru-d-Din Muhammad wrote a short 
grammar of Braj bhasa as part of the introduction to his literary compendium 
Tuhfatu-l-Hmd A present from India,’ written in the Persian language. His 
iutention was to introduce the art of Braj poetry to the Muslim ruling class. He 
took great care in assigning the appropriate Arabic letter to Indian sounds, 
though his phonetic terminology is not scientific; he differentiates d, dh, d and 
dh as ‘lighter d’, ‘heavy d’, ‘d rendering itself heavy’ and ‘heaviest d’. In the 
morphology he always mentions alongside the Persian term its Indian equiva- 
lent and spells it out meticulously. It is interesting that these terms are not in 
Sanskrit form but in a vernacular garb: sandachhar (Sanskrit sandhyahsara 
‘diphthong’), hinjan {vyanjana ‘consonant’), 'purling ['pumlinga ‘masculine’), 
astriliy^ {strilinga ‘feminine’), kirt(*krta ‘object’), etc. The author differentiates 
between those feminine nouns that have a masculine counterpart (e. g. hastanl 
‘female elephant’) and those ‘irregular ones’ that do not (e.g. agan ‘fire’). Many 
instances of old composition he regards as cases of suffixation, e.g. hhup ‘king,’ 
originally a compoimd of hhu ‘earth’ and 'pd ‘rule’ is analyzed as bhu ‘earth’ 
with a suffix -p indicating ‘lordship.’ Another work in the Islamic tradition is 
Insha AUah Khan’s Urdu grammar Darya-e-Latafat (A.D. 1802) written in 

^ H. Saudaota, BSOS 8.718-720. The grammar was not published until 1640. 

2 J.Ph. Vogel, BSOA 8.817-822. 

2 Manoel da Assumpgam possibly was not the sole author of the grammar, but 
edited the work of others (Muhammad A.B. Khondhab, The Portuguese contribu- 
tion to Bengali prose, grammar and lexicography. Doctoral dissertation. University 
of London, 1971, p. 244-246). 

GrrSiinniflirs of tlio Now Indo-Aryaiii Ifiinguages 


The author of the first Marathi grammar is Venkata Madhava, a lecturer of 
Marathi at the Fort St. George College in Madras. His three works on Marathi 
(as it was spoken by the large Maratha colony in Tanjore!) exist only in the 
autographs of the author or his assistant Bhima Pantoa. The Maharastra- 
prayoga-candrika (c. 1827) has 227 sutra-s in Sanskrit and is accompanied by a 
Sanskrit commentary, a Marathi commentary and Marathi illustrations; the 
Sanskrit section is written in Devanagari script, the Marathi in Mo^ script. 
The grammar, which generally follows the Siddhanta Kaumudi in its design, 
was probably meant to introduce Marathi to the neighbouring Tamil speakers.^ 

The Kashmirian I^vara Kaula (1833-1893) wrote his Kasmira-^abdamrta m 
1875 and revised it in 1879. As his object was to describe the Hindu dialect of 
Shrinagar, he eschewed the use of the Arabic script and created a modified 
Devanagari script through the addition of diacritical marks. The phonemes of 
Ka^mM are classified as prasiddha 'known [from Sanskrit]’ and aprasiddha 
'unknown’ ; several phonemes of Sanskrit on the other hand are not found in 
Kai^mM (introduction). The work consists of 778 sutra-s in nine books and a 
commentary in which examples are often translated into Hindi. Isvara Kaula 
follows the terminology and arrangement of the Katantra, with a few interesting 
innovations: e. g. the three persons are characterized as a-srotr ‘non-hearer’ =he, 
srotr 'hearer’ = you, and valctr 'speaker’ = I (VIII 1 3) and the agent of the causa- 
tive is defined as "the imposition of being an agent to somebody else’s action in 
case of default of the proper agent” (VIII 4 1 svdhartrtvdvasare para-kriyd- 
kartrtvdropo hetuh). The Dhatupatha, here rather a list of verb stems than of 
roots (with an additional list of roots that have only nominal derivatives), is 
incorporated into the text as book VII. 

Under the influence of Western linguists a new school of historical and 
descriptive Indian linguists developed early in this century, culminating in the 
founding of the Linguistic Society of India (1928). Soon after the attainment of 
Indian independence, the Language Project at Poona (1954-1959), supported 
by the Hockefeller Foundation, gave the necessary stimulus for expanded 
institutional research in university departments and advanced centres which 
were generously supported by the University Grants Commission. The influence 
of Bloomfieldian and Chomskyan linguistics has occasionally led to a neglect 
of the historical-literary implications and the abandoning of the Indian concept 
of vdg-arthdv iva samprktau 'fused like word and meaning’ (Raghuvamsa II). 
But these problems as well as the achievements of contemporary Indian lin- 
guists are no longer specifically Indian and belong rather in a history of modern 

^ S.D. Labdxt, ABORI 53.260-266. 

® For the more recent development cf. Current trends in linguistics, edited by 
Th. Sebeok, vol. V (The Hague, 1969). 










As. St. 
























JAS (Gale.) 












Annals of the Bhandarkar Oriental Research Institute 

Ail India Oriental Conference 


American Journal of Philology 
Adyar Library Bulletin 
Acta Orientalia (Leiden) 

Abhandlungen der Preussischen Akademie der Wissenschaften 


Asiatische Studien 

Bezzenbergers Beitrage 

Bulletin de Fificole Fran^aise d’Extreme-Orient 

Bharatiya Vidya 


Bulletin of the School of Oriental [and African] Studies 

Bombay Sanskrit and Prakrit Series 


Denkschriften der Kaiserlichen Akademie der Wissenschaf- 
ten, Wien 

Etudes v^diques et panin^ennes 
East and West 
Poundations of Language 

Gottinger Gelehrte Anzeigen 
Giomale della societal Asiatica Italiana 
Indian Antiquary 
Indian Culture 

Indogermanische Porschungen 
Indian Historical Quarterly 
Indo-Iranian Journal 

International J oumal of American Linguistics 
International J oumal of Dravidian Linguistics 
Indian Linguistics 
J oumal asiatique 

Journal of the American Oriental Society 
Journal of the Asiatic Society (Calcutta) 

J oumal of the Bihar and Orissa Research Society 

J oumal of the Department of Letters, University of Calcutta 

J ournal of Indian and Buddhist Studies 

J oumal of Indian Philosophy 

J oumal of the Oriental Institute, Baroda 

Journal of Oriental Research (Madras) 

Journal of the Pali Text Society 

J oumal of the Royal Asiatic Society 

Journal of the University of Poona, reprinted in PC ASS 

J oumal of Vedic Studies 

Kashi Sanskrit Series 






M.6m. Acad, Imp. 







Phil. Ind. 


















(Kuhn’s) Zeitschrift fur vergleichende Sprachforschung 



M^moires d’Academie Imperial St. Petersburg 
M. Monier -Williams, Sanskrit-English Dictionary 
Nachrichten von der Gesellschaft der Wissenschaften, Gottin- 

Now Indian Antiquary 
Oriontalistische Literaturzeitung 

Publications of the Centre of Advanced Study in Sanskrit 
(reprinted from JUP) 

Philosophy East and West 
Philologica Indica (Ps. H. Liiders) 




Sitzungsberichte der Akademie der Wissenschaften, Wien 
Sitzungsberichte der Heidelberger Akademie der Wissen- 

Sitzungsberichte der Preussischen Akademie der Wissenschaf- 



Transactions of the American Philosophical Society 



Vishveshvaranand Indological tToumal 

Wiener Zeitschrift fur die Kunde des Morgenlandes 
Wiener Zeitschrift fur die Kunde Sud[- und Ost]asiens 
Zeitschrift der Deutschen Morgenlandischen Gesellschaft 
Zeitschrift fur Indologie und Iranistik 

The articles by F. Kielhom and P. Thieme have been reproduced with ongmal 
pagination in their “Kleine Sohriften” ; articles of various authors are reprinted mth 
new pagination in J.F. Staal, A Reader on the Sanskrit Grammarians, Cambndge, 
Mass., 1972, 


General Literature: 

Abhyankar, K. V. A Dictionary of Sanskrit Grammar. Baroda, 1961. 

Apte, V.S. Sanskrit-English Dictionary, appendix E, rev. ed. Poona, 1959. 

Belvalkar, S.K. An Account of the Different Existing Systems of Sanskrit Gram- 
mar. Poona, 1915. 

Burnell, A.C. On the Aindra School of Sanskrit Grammarians. Mangalore, 1875. 

Chakravarti, Prabhatchandra. The Linguistic Speculations of the Hindus. Cal- 
cutta, 1933. 

Chatterji, K.C. Technical Terms and Techniques of Sanskrit Grammar, part 1, 
2nd ed. Calcutta, 1964. 

Dandekar, R.N. Vedic Bibliography. Vol. I, Bombay, 1946; vol. II, Poona, 
1961; vol. HI, Poona 1973. 

Haidar, Gurupada. Vyakarana-darsaner itihasa. Calcutta, 1943. 

Palsule, Gajanan Balkrishna. A Concordance of Sanskrit Dhatupathas. Poona, 

Palsule, Gajanan Balkrishna. The Sanskrit Dhatupathas, a Critical Study. Poona, 

Raja, K. Kunjunni. Indian Theories of Meaning, 2nd ed. Madras, 1969. 

Renou, L. Bibliographic v^dique. Paris, 1931. 

Renou, L. La Durghatavrtti de Saranadeva, vol. I, Introduction. Paris, 1940. 

Renou, L. Terminologie grammaticale du Sanskrit, 2nd ed. Paris, 1957. 

Renou, L. and Jakob Wackemagel. Altindische Grammatik, 2nd ed., Introduction 
generale. Gottingen, 1957. 

Ruegg, D.S. Contributions a Thistoire de la philosophie linguistique indienne. 
Paris, 1959. 

Sarma, K. Madhava Krishna. Panini Katyayana and Patanjali. Delhi, 1968. 

Sastri, P.S. Subrahmanya. History of Grammatical Theories in Tamil. Madras, 

Sebeok, Th. A., editor. Current Trends in Linguistics, vol. V. The Hague, 1969. 

Shastri, Haraprasada. A Descriptive Catalogue of the Sanskrit Manuscripts in the 
Collections of the Asiatic Society of Bengal, vol. VI. Vyakarana Manuscripts, 
Calcutta, 1931. 

Tiwari, Bholanath. Bharatiya bha^avijnan ki bhumika. Delhi, 1973. 

Vishva Bandhu. A Vedic Word-Concordance, vol. IV {Vedahga-sutras). Hoshiar- 
pur, 1958-1961. 

Yudhisthir, Mhnamsak. Samskrt vyakaran ka itihas, 3 vols. Bahalgadh, 1973- 

Chapter I: The Origins 

Liebich, B. Zur Einfiihrung in die indische einheimische Sprachwissenschafb II. 
Heidelberg, 1919 (SHAW, no. 15). 

Chapter II: Panini 

Bohtlingk, 0. Panini’s Grammatik, 2nd ed. Leipzig, 1887 (reprint ed., Hrldes- 
heim, 1964). 

A Selected Bibliography 


Renou, L. La grammaire de Panird, 2nd ed. Paris, 1966. 

Vasu, S.C. The Ashtadhyayi of Panini. Benares, 1891-1897 (reprint ed., Delhi, 

Agrawala, V.S. India as known to Panini. Lucknow, 1953. 

Bahulikar, S.D. Some Criteria for Determining the Insertions in the A^tadhyayL 
Doctoral dissertation, Harvard University, 1972. 

Birw^, R. Der Ganapatha zu den Adhyayas IV und V der Grammatik Paninis. 
Wiesbaden, 1961. 

Birwe, R. Studien zu Adhyaya HE der A§tadhyayi Paninis. Wiesbaden, 1966. 

Buiskool, H.E. Purvatrasiddham. Amsterdam, 1934. 

Buiskool, H.E. The Tripadi. Leiden, 1939. An abridged English recast of Pur- 

Cardona, G. The Method of Description reflected in the Sivasutras. Studies in 
Indian Grammarians, I; Transactions of the American Philosophical Society, 
no. 59, 1. Philadelphia, 1969. 

Devasthali, G.V. Anubandhas of Panini. Poona, 1967. 

Faddegon, B. Studies on Panini’s Grammar. Amsterdam, 1936. 

Goldstiicker, Th. Panini, his Place in Sanskrit Literature. London, 1861 (reprint 
ed., Osnabriick, 1966). 

Katre, S.M. Paninian Studies, vol. I-IH. Poona, 1967-1969. 

Liebich, B. Panini. Leipzig, 1891. 

Misra, V.N. The Descriptive Technique of Panini. The Hague, 1966. 

Pathak, Shridharshastri and Siddheshvarshastri Chitrao. Word Index to Panini- 
sutra-patha and Pari^i^tas. Poona, 1935. 

Pawate, I.S. The Structure of the Ashtadhyayi. Ho date [Hubli, 1935(?)]. 

Rocher, R. La th^orie des voix du verbe dans F^cole Panin6enne. Bruxelles, 1968. 

Scharfe, H. Panini’s Metalanguage. Philadelphia, 1971. 

Sen, Sukumar, Paninica. Calcutta, 1970. 

Shefts, B. Grammatical Method in Panini: His Treatment of Sanskrit Present 
Stems. Hew Haven, 1961. 

Staal, J.F. Euklides en Panini: twee methodische richtlijnen voor de filosofie. 
Amsterdam, 1963. 

Thieme, P. Panini and the Veda. Allahabad, 1935. 

Wezler, A. Bestimmung und Angabe der Funktion von Sekundar-Suffixen durch 
Panini, Wiesbaden, 1975. 

Chapter IH: Yaska 

Bhadkamkar, H.M. and R.G. Bhadkamkar, editors. The Hirukta of Yaska, with 
Durga’s commentary. Bombay and Poona, 1918-1942. 

Rajavade, V.K. Yaska’s Hirukta: Text and Exegetical Hotes. Poona, 1940. 
Sarup, Lakshman, editor and translator. The Highantu and the Hirukta. Lahore, 
1927 (reprint ed., Delhi, 1967). Quoted by book and chapter. 

Sarup, Lakshman, editor. Hirukta, with the commentary of Skandasvamin and 
Mahe^vara. Lahore, 1928-1934. 

Bhattacharya, Bishnupada. Yaska’s Hirukta and the Science of Etymology. 
Calcutta, 1958. 

Macdonell, A. A., editor and translator. The Brhad-devata attributed to Saunaka. 

Cambridge, Mass., 1904 (reprint ed., Delhi, 1965). 

Shastri, Pt. Shivanarayan. Hirukta-Mimamsa. Benares, 1970. 

Skold, H. The Hirukta. Lund, 1926. 

Varma, Siddheshwar. The Etymologies of Yaska. Hoshiarpur, 1953. 


H. Scliarfe • Grammatical Literature 

Chapters IV and VT—^THi Katyayana, Patanjali, etc. 

Chatterji, K.C. Patanjali’s Mahabha^ya. Calcutta, 1957. 

Joshi, Bharga^-asastrin, Sivadatta B. Kudala and Raghunatha^astrin, editors. 
Vyakarana-Mahabha^ya. Bombay, 1935-1951. Includes also Kaiyata’s Pradipa 
and Nagojibhatta’s Uddyota. Anon. repr. ed. New Delhi 1967. 

Joshi, S.D. and J.A.F. Boodbergen, editors. Patahjali’s Vyakarana-Mahabhasya 
(with translation and notes). Poona: Samarthahnika, 1968; Avyayibhavatat- 
puru^ahnika, 1969; Karmadharayahnika, 1971; Tatpurusahnika, 1973. Bahu- 
vrihidvandvahnika 1974. 

Kielhora, F., editor. The Vyakarana-Mahabha§ya. Bombay, 1880-1885. 3rd 
edition revised by K.V. Abhyankar, Poona, 1962-1972. Quoted by volume, 
page and line. Repr. ed. of 2nd ed. Osnabriick 1970. 

Pathak, Shridharshastri and Siddheshvarshastri Chitrao. Word Index to Patah- 
jali’s Vyakarana-Mahabha^a. Poona, 1927. 

Sastri, P.P.S., A. Sankaran and T. Chandrasekharan, editors. The Vyakarana- 
Mahabhasya, parts 1 and 2. Madras, 1948-1952. Also includes Kaiyata’s 
Pradipa and Annambhatta’s Mahabhasya-pradipoddyotana. Only goes up to 
Panini I 1 74. 

Sastri, P.S. Subrahmanya. Lectures on Patanjali’s Mahabha?ya. Annamalai, 
TiruchirapaUi and Thiruvaiyaru, 1943-1962. The English translation of 
Adhyaya-s I and II. 

Veda\Tata, editor. Vyakarana-Mahabhasya. Rohtak, 1962-1963. Also includes 
Kaiyata’s Mahabha^ya-pradipa and Nagojibhatta’s Bha^ya-pradipoddyota. 

Insler, S. Verbal Paradigms in Patanjali. University Microfihna, Ann Arbor, 
1967 (Yale thesis, 1963). 

Kjelhom, F. Katyayana and Patanjali: Their Relation to each other and to 
Panini. Bombay, 1876 (reprint ed., Benares, 1963). 

Lahiri, P.C. Concordance Panini — Patanjali. Breslau, 1935. 

Limaye, V.P. Critical Studies in the Mahabha^ya. Hoshiarpur, 1974. 

Paranjpe, V. G. Le vartika de Katyayana, une 4tude du style, du vocabulaire et 
des postulates philosophiques. Heidelberg, 1922. 

Puri, B.N. India in the Time of Patanjali. Bombay, 1968. 

Sarma, K. Madhava Krishna. Panini Katyayana and Patanjali. Delhi, 1968. 

Scharfe, H. Die Logik im Mahabha^ya. Berlin, 1961. 

Thieme, P. Bh^ya zu Varttika 5 zu Panini 1.1.9 und seine einheimischen Erkla- 
rer. Gottingen, 1935 (NGGW N. F. 1, no. 5). 

Wezler, A. Paribha§a IV, V und XV. Bad Homburg, 1969. 

Chapter V : The Prati^akhya-s 

Burnell, A.D., editor. Rktantra-vyakarana. Mangalore, 1879. 

Rastogi, Shrimati Indu, editor and translator. Katyayana; Suklayajurveda- 
prati^akhyam. Benares, 1967. 

Sarma, V. Venkatarama, editor. Taittiriya-prati^khya, with Mahiseya’s Bha^ya. 
Madras, 1930. 

Sarma, V. Venkatarama, editor. Vajasaneyi-prati^akhya. Madras, 1934, 

Shastri, Mangal Deva, editor and translator. Rgveda-prati^akhya. Vol. I, Bena- 
res, 1959; vol. II, Allahabad, 1931 ; vol. HI, Lahore, 1937. 

Shastri, Surya Kanta, editor. Rktantram. Lahore, 1933 (reprint ed., Delhi, 1970). 
Shastri, Surya Kanta, editor. Atharva Pratisakhya. Delhi, 1968. 
Taittiriya-prati^akhya with Uwata’s Bha§ya. Benares Sanskrit Series 5. 

Varma, Virendra Kumar, editor. Rgveda-prati^akhya, with Uvata’s commentary. 
Benares, 1970. 

A Selected Bibliography 


Weber, A., editor. Vajasaneyi>prati9akhya of Katyayana. Indische Studien 4, 
65-171; 177~-331. 

Whitney, W.D., editor and translator. The Atharva-veda Prati^akhya or Sauna- 
kiya Catnradhyayika. JAOS 7. 333-615 (reprint ed., Benares, 1962). 

Whitney, W.D., editor. Taittiriya Prati^akhya. JAOS 9. 1-469 (reprint ed., Delhi, 

Allen, W. S. Phonetics in Ancient India. London, 1953 (reprint ed. Benares, 1962), 
Gelpke, F. Anantabhatta’s Padarthapraka^a: Ein Kanva-Konunentar zum Vaja- 
saneyi-prati^akhya. Gottingen, 1929. 

Sarma, V. Venkatarama. Critical Studies on Katyayana’s Sukla-Yajurveda- 
Prati^akhya. Madras, 1935. 

Varma, Siddheshwar. Critical Studies in the Phonetic Observations of Indian 
Grammarians. Oxford, 1929 (reprint ed., Delhi, 1961). 

Varma, Virendra Kumar. Rgveda-pratisakhya; ek anu^ilan. Benares, 1972. 

Chapters VT— VIII, see Chapter IV. 

Chapter IX: The Buddhist Sanskrit Grammarians 

Chatter ji, K.C., editor. Candra Vyakarana of Candragomin. Poona, 1953-1961. 

Contains sutra-s, vrtti, varna-sutra-s and paribha§a-sutra-s. 

Eggeling, J., editor. Katantra. Calcutta, 1874-1878. Includes commentary by 

Liebich, B., editor. Candravrtti. Leipzig, 1918 (reprint ed., Wiesbaden, 1966). 
Liebich, B., editor. Candra-vyakarana, die Grammatik des Candragomin. Leip- 
zig, 1902 (reprint ed., Wiesbaden, 1966). 

Liebich, B., editor. Das Katantra, Zur Einfuhrung in die indische einheimisohe 
Sprachwissenschaft I. Heidelberg, 1919 (SHAW no. 4). 

Liebich, B. Ztit Einfuhrung TV, Analyse der Candra-Viiiti. Heidelberg, 1920 
(SHAW no. 13). 

Liebich, B. Konkordanz Panini — Candra. Breslau, 1928. 

Renou, L. Les “innovations” de la grammaire de Candragomin. Etudes de gram- 
maire Sanskrite, 3. Paris, 1936. 

Chapter X: The Jain Sanskrit Grammarians 

Jaina, Srilala, editor. Sabdarnavacandrika. Benares, 1915. 
Siddhahem^abdanu4asana with Laghuvrtti. Benar^, 1905. 

Suri, Vijayalavanya, editor, ^risiddhahemacandra Sabdanusasanam with Brhad- 
vrtti and Kyasa. Bombay, 1960. 

Tripathi, Sambhunath, editor. Jainendra Vyakaranam . . . with Mahavrtti. Bena- 
res, 1956. -ct r 1, 

Tripathi, Sambhunath, editor. Sakatayana-vyakaranam. Benares, 1971. EnglisH 

introduction by R. Birwe. 

Chapter XI: The later Panini School 

Abhyankar, K.V., editor. Paribha§endu4ekhara of Nagojibhatta. Poona, 1962. 
Abhyankar, K.V. and V.P. Limaye, editors. Bhairtrhari’s Mahabha^yadipika. 

ABORT 43-50; Poona, 1970. ^ • -d 

Abhyankar, K.V. and V.P. Limaye, editors. Vakyapadiya of Bhartrhan. Poona, 


Apte, Hari Narayana, editor. Paribha§endu^kliaxa with Vaidyaaatha Paya- 
^da’s eoromentary. Anandairama Sanskrit Series 72, 1913. 


H. Scharfe • Grammatical Literatiare 

Tolkappiyam, CoUatikaram; with Cenavaraiyar’s commentary. Madras, 1959. 

Tholkappiyam (in English) with Critical Studies by S. Haldmvanar. Madurai, 1963. 

Tolkappiyam, with Ilampuranar’s commentary. Madras, 1955—1963. 

Tolkappiyam, with Naccinarkkiniyar’s commentary. Madras, 1944-1965. 

AkattiyaJihkam, C. and K. Murukaiyan, editors. Tolkappiya Moliyiyal. Anna- 
malainagar 1972. 

Aiyar, L. V. Ramaswami. Grammar in Lilatilakam. Trichur, 1944. 

Pillai, Vaiyapuri S. History of Tamil Language and Literature. Madras, 1956. 

Reddiar, Venkatarajulu. Tolkappiya eluttatikara-v-araycci. Madras, 1944. 

Sastri, P.S. Subrahmanya. History of Grammatical Theories in Tamil. Madras, 

Sastri, P.S. Subrahmanya. An enquiry into the relationship of Sanskrit and 
Tamil. University of Travancore, 1946. 

Shanmugam, S.V. Haccinarkmiyar’s Conception of Phonology. Annamalainagar, 

Vellaivaranan, K. Tolkappiyam-Hannul eluttatikaram. Annamalainagar, 1962. 

Vijayavenugopal, G. A modem Evaluation of Hannul (eluttatikaram). Annama- 
lainagar, 1968. 

Chapter XIV : Other Systems of Sanskrit Grammar 

Bohtlingk, 0., editor. Vopadeva’s Mugdhabodha. St. Petersburg, 1847. 

Kaviratna, Syamacarana, editor. Sainkjjiptas^a of Kramadilvara. Calcutta, 

Muni, Jina Vijaya, editor. Ukti-Vyakti-Prakarana of Pandita Damodara. Bom- 
bay, 1953. 

Sarma, Hava Kishora Kara, editor. The Sarasvata Vyakaranam of Anubhuti 
Svarupacharya. Benares, 1935-1936 (reprint ed., 1971). 

Sastri, K. Sambasiva and V.A. Ramaswami Sastri, editor, Saraavatikant^habha- 
rana of Sri Bhojadeva. Trivandrum, 1935—1948 (incomplete). 

Chapter XV: Grammars of the Middle Indo -Aryan Languages 

Cowell, E.B., editor and translator. PrakrtaprakaAaofVararuchi. Hertford, 1854; 
reprint ed., Calcutta, 1962; revised edition by P.L. Vaidya, Poona, 1931. 

Ghosh, Manomohan, editor. Prakrtakalpataru of Rama^arman. Calcutta, 1954. 

Gnanaloka, Kodagoda, editor. Sidat-sahgara. Weligama, 1971. 

Hoemle, A.F.R., editor. Prakrta-lakshanam or Chanda’s grammar. Calcutta, 1880. 

Hausalyayana, Bhadanta Ananda, editor. Moggallana-vyakarana. Hoshiarpur, 

Nitti-Dolci, L., editor. Prakrtanu^asana de Puru§ottama. Paris, 1938. 

Pischel, R., editor. Hemacandra; Grammatik der Prakritsprachen. HaUe, 1877- 

Raja, C. Kxmhan and H. Ramachandra Sarma, editors. Prakrta-praka4a of 
Vararuci with the commentary of Ramapanivada. Madras, 1946. 

Smith, H,, editor. Saddaniti; la grammaire palie d’Aggavarnsa. Lund, 1928-1966. 

Sthavira, Sri Dharmananda Hayaka, editor. Moggallana Pancika. Wirahena 

Tiwari, Lakshmi Xarayan and Birbal Sharma, editors. Kaccayana-Vyakarana. 
Benares, i962. 

Upadhyay, Balladeva, editor. PrakrtaprakaAa of Vararuci. Benares, 1972, 

Vaidya, P.L., editor. Prakrit Grammar of Trivikrama. Sholapur, 1954. 

Vaidya, P.L., editor. Hemacandra’s Prakrit Grammar. Poona, 1958. 

A Selected Bibliograpliy 


Franke, R.O. Geschichte und Elritik der emheimischen PaK-Grammatik und 
-Lexicographie. Strassburg, 1902. 

Geiger, W. Pali, Literatur und Sprache. Strassburg, 1916 [Engl, transl. Calcutta, 
1943; repr. ed. Delhi, 1968]. 

Nitti-Dolci, L. Les grammairiens prakrits. Paris, 1938. 

Pischel, R. Grammatik der Prakrit-Sprachen. Strassburg, 1900. 

Chapter XVI: The Parasi-prakasa 

Srivibhutibhu^anabhattacarya, editor. Parasipraka^a by Bihari Kr§na Dasa 
MMra. Benares, 1965. 

Weber, A., editor and translator. "Cber den zweiten, graimnatischen, Parasi- 
prakaga des Krishnadasa. Berlin, 1889 (APAW). 

Chapter XVII: Grammars of New Indo -Aryan Languages 

"Abd al-Haqq, editor. Insha Allah Elhan: Darya i Latafat. Lucknow, 1916. 

Abdul Ra‘uf ‘Aruj, translator. Darya i Latafat. Karachi, 1962. 

Arjunwadkar, K.S., editor. VemkatamadhavakrtaMahara^tra-prayogacamdrika. 
Poona, 1970. 

Grierson, G. A., editor. Ka 5 mira 9 abdamrta by I^vara-Kaula. Calcutta, 1897-1898. 
Ziauddin, M., editor and translator. A Grammar of the Braj Bhakha by Mirza 
Khan (1676 A.D.). Calcutta, 1935. 


X. Indian Terms 

akalaka 114 
alrtinai 181 
ak$ara 77 

aksara-samamnaya 78 
aiiguli-sphotana 177 
acai-ccol 121 fh. 22 
acai-nilai 121 fn. 22 
adhikarana 94; 98 
adhikara 94 
anun^ika 112 
anubandha 145 fn. 48 
anuvrtti 87; 159 
anusvara 82 
antaranga 145 f.; 158 
antastha 78 
anyatarasyam 166 
anyade^ya 186 
anvaya 148 

apabhram^a 184 fn. 36; 188; 193f. 

apavada 145; 158 

apad^a 94; 98; 173 

apaya 160 

aprasiddha 199 

abhinihita-samdhi 107f. 

avagraha 127 

avasana 78; 108 

avidya 172 

a^otr 199 

asamjnaka 165f. 

asiddha 101 ; 141 

akrti 124 

akrti-gana 102 

akhyata 84; 103 fn. 69; 112 fn. 104; 119 
fn. 12; 121 

agama 130; 147 fn. 59; 160 
acarya>desiya 152 fn. 3 
Mesa 109f.; 140; 156; 160 
adhara 94; 166 
aytam 180 
ardhadhatuka 85 
aryavarta 107; 153; 156f. 
arsa 162; 193 
ascarya 145 
it 93 fh. 29 
iti 93 fn. 28; 138 

ukti 188 
uccaranartha 90 

nn-Mi 104f.; 113; 119; 163f.; 169; 187 

uttara-pada-lopa 147 

utsarga 145; 158 

udicam 108 

udgraha 127 

upakara 173 

upakaraka 173; 182 

upajana 160 

upade^a 93; 192; 196 fn. 2 
upasarga 85 f.; 121 
uyartinai 181 
uyir 179 fn. 15 
u§man 78 
eka-tih 148 
eka-^e§a 136; 168 
eluttu 178; 180 fn. 16 
aintiram 179 
aupacarika satta 172 
karana 94 f.; 173 
karisyat 80 

kartr 94f.; 127; 163 fn. 7; 195 
karman 94f.; 127; 155; 160; 173; 181 
-kara 78 

karaka 94-100; 112; 127; 165f.; 169; 

173; 182f.; 186; 189; 194f. 
karitam 182 
karya 125 
knrvat 80 
kurippu 181 f. 
kr 127; 181 

krt 97-99; 110; 118; 127 
krta 80 

kriya 80 fn. 23; 97 fn. 45; 119-121; 155; 
175 fn. 30 

gana 102-104; 108 fn. 91; 133; 144; 169; 

gana-sutra 103 
gnna 111; 121; 163 
guna-vacana 120 fn. 16 + 19 
guna-^abda 120 fn. 19 
grh 127 
grahana 127 
gramya 186 



gho§avat 129 

cakrvas 80 fn. 22 

cekriyita 86 fn. 46 

cey 181 

col 178 

chandas 106 

tatsama 184; 186 

taddhita 100; 110; 118; 151; 182 

tadbhava 184; 186 

tantra 87 

devanampriya 138 

de^ya 184; 186 

dyotaka 85 

dravya 120f. ; 124; 136 
dhatu 84f.; 96; 121; 182 
dhvani 172 

naman 112 fn. 104; 119 fn. 12; 120 

nitya 110; 125 

nipatana 103; 108 

nirukta 78; 82-84; 117 

nirudha-laksana 110 

nirvacana 83 

nivrtti-sthana 118 

noti 177 

nyaya 169 

panca-varga 82 

pada 77; 81 

paribha^a 93; 164 

pa^yanti 172 fn. 20 

pada 77 

parsada 127 

pal 181 

purling 198 

purvacaxya 86 fn. 46 

purva-vipratisiddha 145 

porul 178 

pragrhya 106; 144 

pragraha 127 

pratyahara-sutra 92 fn. 24 
prade^a 121 f. 
prasiddha 199 
prakrta 191 
pracam 108 
prade^ika 122 
phakkika 143 
phala 161; 175 fn. 30 
bala 79 

bahiranga 146; 158 
binjan 198 
bindu 162 
brahman 172 
brahma-ra4i 161 
bhavat 80 
bhavanti 140 

bhavisyat 80; 97 

bhava 80 fn. 23; 97 fn. 45; 119f.; 155 

bhavini samjha 161; 173 

bha?a 105f.; 184 

bha§ya 152 

bhuta 80; 97 

bhuv-adayah 96 

bheda 126 

mani-pravala 183-185 
madhyama-pada-lopa 147 
madhyama 172 fn. 20 
matra 79 

mahe^vara-sutra 92 fn. 24 
mukhya satta 172 
mey 179 fn. 15 
mleccha 157 
lipi 113 

lopa 93; 147; 160 
vaktr 199 
-vat 81 

varga 100 fa. 58 

varna 78f.; 115; 164 

varna-vyatyaya 118 fn. 7 

vartamana 80; 97 

varti(n) 126 

va 166 

vag-yoga 172 

vac 77; 172 fn. 20 

varttika 135 

vikarana 96f.; 155f. 

vikara’l22; 130; 140; 160; 180 

vinai 181 

viprati^edha 145; 166 
vibhakti 80; 98; 168 
vibha§a 166 
vivak§a 172 
visarga 82; 162 
vrddhi 111; 163; 165 
vedahga 78; 83 fn. 32; 108; 172 
verrumai-mayakkam 181 
vaikhari 172 fn. 20 
vaidhaveya 160 
vyanjana 198 
vyatireka 148 

vyakarana 80 fa. 20; 82f.; 112 
vyakr 80 

vyapara 175 fh. 30 
^akti 172 

^bda-nityatva 86 fa. 47; 110 
sabda-purva-yoga 172 
4aka-parthiva 138 
^abda-bodha 175 fii. 30 
salaturiya 88 
sik§a 78f.; 82; 85; 176f. 


H. Scharfe • Grammatical Literature 

siva-sutra 92 fn. 24 

^§ka-tarka 171 

se§a 95 fn. 36 

^otr 199 

^astani 140 

sarosarga 124 

samskrta 157 fn. 25 

samhita 81; 100 fn. 58; 118; 127f. 

samhita-patha 81; 89; 127 

samjna 93 

satta 172 

sattva 119->121 

samtana 79 

sandachhar 198 

samdhayaka 185 

samdhi 79 fh. 14; 81; 98; 101; 127; 130; 

180; 183 

samdhy-ak?ara 103 fn. 69; 128; 198 
sampurna-gana 102 
sampradana 94; 173 
samprasarana 111 

samartha 99 
samahara 92 
savarna 112; 129; 140 
sadhana 172 
saman 79 

sarvadhatuka 85; 96 

siddha 125; 135; 141 

siddhanta 152 

suta 157 

Sutra 86 f. 

sutra-bheda 143 

sthana 82; 93 

sparddha 166 

spar4a 78 

spar^agho§a 140 

sphota 172 

sphotana 132 

svara 78f. ; 128; 140 

svara-bhakti 90 fn. 17; 131 

hetu 199 

B. Indian Authors 

Agastya 179 
Aggavamsa 194f. 

Atharvanacarya 185 
Anantabhatta 134 
Anubhutisvarupacarya 163; 189 
Appakavi 185 
Abhayanandin 168 
Abhinavagupta 172 
Amaracandra 95 fn. 36 
AhobaJapati 85 
Atreya 134 
Api^aii 119 fh. 11 
Insha Allah Khan 198 
Ilampuranar 179 
Kvara Kaula 199 
Ugrabhuti 117 fn. 4 
Ujjvaladatta 104 fn. 76 
ITvata 133f. 

Rishikesh Sastri 194 
Elakuci Balasarasvati 185 
Audumbarayana 86; 172 
Kaccayana 163; 194f. 

Katya 135 fn, 1 ; 149 fn. 2 
Katyayana 82 fn. 29; 83 fn. 32; 88 f.; 

91; 95; 110; 113; 114 fn. 116; 118 fn. 

7; 124; 129; 131; 135-150; 152f.; 

157f.; 160-162; 168; 173 fn. 21; 174 
Kasakxtsna 163 fn. 5 
Kaj§matha 190 

Kaiyapa 164 fn. 14 
Kunavira Pantita 183 
Kum^alata 162 
Kumariia 185 fn. 40 
Kr^nadasa 183 fn. 31; 196f. 
Krsnamacarya 186 
Ketana 185f. 

Kesiraja 186 

Kaiyata 149; 170 fn. 3; 175 
Kautsa 122 

Kramadi^vara 164 fn. 12; 193; 195 
K§irasvamin 101 fn. 63 
Gargya 85f.; 109 fn. 93; 119; 121 
Gonardiya 154 fn. 12 
Gonikaputra 154 fn. 12 
Goyicandra 188 fn. 3 
Canda 194 

Candragomin 84; 102 fn. 66; 114; 154; 
163-168; 174 fn. 24; 182 fn. 28; 194 
fn. 8; 195 

Cuppiramaniya-titcitar 183 
Cenavaraiyar 179 
Jayaditya 174 
Jinendrabuddhi 104; 174 
Jivagosvamin 189 fn. 6 
Jumaranandin 188 
Teyvaccilaiyar 179 
Tolkappiyan 178—181 
Trivikrama 193f. 



Damodara 188 
Durgasimha 1) 163 2) 117f. 
Devanandin 112; 168 f. 

Dharmakirti 174f.; 189 
Dharmadasa 165 
Naccinarkiniyar 179 
Nannaya [bhattu] 185 
Nagavarma 186 

Nagojibhatta (Nagesa) Kale 91 fn. 20; 
137 fn. 15; 141f.; 143 fn. 40; 146; 
159; 170 fn. 3; 175 
Narayanabhatta 1) 187 2) 174 
Narayana Vidyavinoda 192 fn. 2 
Nilakantha 117 fn. 4 
N^^atunga 186 fn. 44 
Patanjali 84; 89; 98; 102; 105; 108; 118; 

125; 137-139; 152-161 
Padmanabhadatta 190* 

Panamparanar 178f. 

Paravastu Chinnayasiiri 185 
Pavananti 183 

Panini 80f.; 84-86; 88-116; 118-120; 
128f.; 132; 140f.; 147; 152; 157; 159; 
168; 195 
Palyakirti 169 
Punyaraja 171 
Puttamittiran 182f. 

Puru^ottama 1) 174; 192 2) 190 
Pujyapada 168 
Peruntevanar 182 
Peraciriyar 179 
Pauskarasadi 144 
Phuilaraja 171 

Buddhapriya Dipamkara 194 
Buddhisagara Suri 169 fn. 10 
Baiji 171 

Bhattakalanka Deva 186 
Bhattoji i)ik§ita 102; 174; 187 
Bhartrhari 170-175 
Bhamaha 192 
Bharadvajiya 149f. 

Bhimasena 102 

Bhoja 169; 183 fn. 33; 187 

Mayilainatar 183 

Malay agiri 169 

M&rkandeya 192 

Mahi§eya 134 

Mirza Khan 198 

Maitreyaraksita 101 fn. 63; 195 

Moggallana 166; 195 
Mohadeva Vedantin 104 fn. 76 
Yaska 83f.; 117-123; 147 
Rajarajavarman 185 
Ramacandra 174f. 

Ramacandra^rama 189 fn. 8 
Ramapanivada 192 fn, 2 
Ramaiarman 192 
Rupagosvamin 189 fn. 6 
Varadaraja 190 
Vararuci 134f.; 162; 192 
Vasantaraja 192 fn. 2 
Vasurata 126; 170 
Vakpati 191 

Vajapyayana 124-126; 137f. 

Vamana 1) 174 2) 102 
Vitthala 91 fn. 20 
Vimaiabuddhi 194 
Vimalasarasvati 174 
Visnumitra 133 
V^^abhadeva 171 
Venkata Madhava 199 
Vedamitra Sakalya 81 
Vedeha Thera 195 
Vaittiyanata Kavalar 183 
Vopadeva 112; 163; 188f. 
Vyaghrabhuti 149 
Vyadi 124-126; 131; 136-138 
Sabarasvamin 142 
Sarvavarman 162-164; 175 
Sakalya 192 £n. 3 

Sakatayana 85 f.; 104; 109 fn. 93; 119; 

121; 132; 162; 169 
Santanava 105 
Saunaka 132 
Sadananda 192 fn. 2 
Sadhnsundara Gam 188 fn. 4 
Sayana 101 fn. 63; 114; 133 
Sitaramaeajyulu 186 
Somadeva 168 
Saunaga 149 
Saubhava 171 
Skandasvamin 117 fn. 4 
Haradatta 174 
Haryak^a 171 

Hemacandra Suri 163f.; 169; 188; 191; 

Hel^aja 124f.; 171 


H. Scharfe • Grammatical Literature 

C. Indian Texts 

Akattiyam 179 fn. 9 
Atharvaveda-pratisakhya 130 fn. 10; 

Amogha-vrtti 169 
Amoghanandini-siksa 177 
Arthasastra 120; 188 
Avinayam 182 

A?tadhyayi 88-116; 174f.; 194 
Andhra-bha§a>bhu§ana 185 
Andhra-i^abda-cintamani 185 
Apisali-iik§a 176 
Ilakkana-vilakkam 183 
Ukti-ratnakara 188 fn. 4 
Ukti-vyakti-prakarana 188 
Unadi-kosa 104 fn. 76 
Unadi-sutra 104f.; 119; 147; 164 
Rktantra 127 f.; 133 
Rgveda 90 fn. 14; 116; 161 
Rgveda-prati^akhya 91 fn. 20; 92 fn. 25 ; 
104; 109 fn. 93; 120f.; 125; 127f.; 

Aitareya-brahmana 77f.; 80 f.; 84 
Aitareyaranyaka 81 
Aindra-vyakarana 179 
Kalapa 162 

Karnataka-bhasa-bhusana 186 
Karnataka-^abdanu^asana 186 
Kavirajamarga 186 fn. 44 
Kateira-^abdamrta 199 
Katantra 162f.; 169; 174; 188; 195 
Karaka-nirupana 95 fn. 36 
Kavyavalokana 186 
Kasika-vivarana-panjika 174 
Ka^ika-vrtti 89; 102; 108; 165f.; 168; 

174; 194 fn. 8 
Kerala-paniniyam 185 
Kaumara 162 
Kaumaralata 162 
Kausitaki-brahmana 106 
K§ira-tarangini 101 fn. 63 
Gana-patha 102; 115 fh. 120; 121 
Girvana-pada-manjari 190 
Gautaml-^iksa 176 

Caturadbyaj^a 109 fn. 93; 130; 134 
Candravrtti 114; 164-166 
Chandogyopanisad 77 fh. 2; 78 
Jainendra-vyakarana 82 fh. 29; 112; 

166; 168f.; 186 
Tuhfatu-l-Hind 198 

Taittiriya-pratisakhya 127f.; 133f.;177 
Taittiriya-sambita 80 
Taittiriyaranyaka 78 

Tonnul-vilakkam 183 
Tolkappiyam 177-181; 183; 185 
Trikandi 171 

Tripadi 1) 101; 141; 145 2) 170 
Tribba§ya-ratna 134 
Trilihga-^abdanusasana 185 
Darya-e-Latafat 198 
Dhatu-kavya 174 fn. 26 
Dbatu-patba 84; 94; 96; 101-104; 107; 
163f.; 187; 199 

Dbatu-pradipa 101 fn. 63; 195 
Dbatu-mala 195 

JSTannul 87 fn. 49; 177; 179; 183; 185 
Narada-^iksa 176 
Mgbantu 117 

Nirukta 84f.; 117-123; 127 fn. 1; 181 
Nirukta-bba§ya 117 fn. 4 
Nirukta-^lokavarttika 117 fn. 4 
Neminatam 183 
Nyasa 1) 104; 114; 174 2) 194 
Patanjali-carita 154 
Pada-patba 81 ; 127 
Pada-manjari 114; 174 
Paribba§a-sucana 124 fn. 4 
Paribha§endu^ekbara 175 
Paniniya-iSik^a 176 
Parasi-prakasa 182fh.31; 196 f. 
Pari-^iksa 176 
Par§ada-vrtti 133 
Par^ada-vyakbya 133 
Pirayoka-vivekam 183 
Prakirnaka 171 
Prakirnaka-praka^a 171 
Prakirnaka-vivarana 172 fn. 18 
Prakriya-kaumudi 174 
Prakiiya-sarvasva 174 
Pradipa 1) See Mababba^ya-pradipa 2) 

Prayoga-ratna-mala 190 
Prakrta-kalpataru 192 
Prakrta-pada 192 fn. 2 
Prakrta-praka^a 192 
Prakrta-manjari 192 fii. 2 
Prakrta-lak§ana 194 
Prakrta vyakarana 194 
Prakrta-^abdanu^asana 193f. 
Prakrta-samjivani 192 fn. 2 
Prakrta-sarvasva 192 
Prakrta-subodbini 192 fn. 2 
Prakrfcanui^asana 192 
PrMkkbya 127-134; 140f.; 145; 163; 



Praudha-manorama 174f. 
Praudha-vyakaranamu 186 
Phit-sutra 105 
Bala-manorama 174 
Bala-vyakaranamu 185f. 
Balavabodhana 164 fn. 14 
Buddhisagara-vyakarana 169 fn. 10 
Brhacchabdendusekhara 175 
Brhaddevata 104; 118; 120 
Brhadvrtti 169 
Brhannyasa 169 
Brhanmanju^a 175 fn. 30 
Bharata-natya^astra 1 9 1 f . 

Bha$av:rtti 174 

Mahabha§ya 84; 107 f.; 120; 126; 130; 

135; 149-161; 164; 174f.; 181 
Mahabhasya-dipika 125; 161; 170 f. 
Mahabhasya-pradipa 170 fn. 3; 175 
Mahabha§ya-pradip6ddyota 175 
Mahara^tra-prayoga-candrika 199 
Magadha-sadda-lakkhana 195 
Mandavi4ik§a 177 
Madhaviya-dhatuvrtti 101 fn. 63 
Mimamsa-sutra 92 fn. 25; 138 
Mngdhabodha 112; 188f. 
Mugdhavabodha-mauktika 188fn. 4 ; 193 
fn. 5 

Yajurveda, K;^na 89; 106f.; 119; 176 

Yajurveda, gukla 89; 106; 119; 129; 134 

Yoga-sutra 154 

Rasavati 188 

Rupamala 174 

Rupasiddhi 194 

Rupavatara 174 

Laghuvrtti 169 

Linganu^sana 1) 105 2) 163 3) 169 
Lilatilakam 183-185 
Lekba-paddhati 188 fn. 5 
Vararuci-karika 182 
Vargadvaya-vrtti 133 

Vakyapadiya 119f.; 125f.; 154; 164 fn. 

9; 170-174; 191 
♦Vakyapradipa 171 

Vajasaneyi-pratilakhya 121; 127-130; 
132-134; 140 

Varttika 135-151; 174; 187 
Vikrti-viveka 185 

Viracoliyam (or Viracoliya-kkaxikai) 
173; 182f.; 195 

Vaiyakarana-siddhanta-nianju§a 175 
Vyasa-sik§a 177 
Sakra-vyakarana 179 fn. 7 
Satapatha-brahmana 77 
^abda-dhatu-samik§a 171 
Sabda-prabha 171 
Sabda-mani-darpana 186 
Sabda-smrti 186 

Sakatayana-vyakarana 82 fn. 29; 166; 

Sankhayanaranyaka 80 
Slokavarttika 149 
Samhita-patha 127 
Samk§ipta-sara 187f.; 193 
Samgraha 124f. 

Sadda-niti 194f. 

Sarasvati-kanthabharana 169; 187; 189 
Saras vati-prakriya 189 
Sarasvata-vyakarana 189f. 
Sidat-sangarava 195 
Siddhanta-kaumudi 114; 164 fn. 12; 
174; 185 

Siddh^ta-candrika 189 fn. 8 
Siddhahaimacandra 169; 193 
Subrahmanya 141 
Saupadma-vyakarana 190 
Svara-vyanjana-siksa 176 
Harinamamrta 189 fn. 6 
Hrdaya-harini 187 
Hosagannada nudigannadi 186 

D. Miscellany 

ablaut 11 Of. 

accents 89; 94; lOOf.; 108; 132; 141; 

164; 168; 187; 189 
active voice 97f.; 155 
adjective 120; 126; 181 
allophones 82 
Arabs 79 f. 

Beschi, C. J. 183 
Bohtlingk, O. 114f. 

Buddhist Sanskrit 162 

case endings 95 f.; 166; 180f.; 185; 192; 
194; 196 

Chinese 79; 167 fn. 22 
Colebrooke,H.Th. 114 
composition 99; 156; 181 
conversation manual 190 
Denkart 80 fn. 20 

determinatives 90; 93; 119; 145; 160; 

163; 166; 189; 194 
direct method 190 


H. Scharfe * Granmiatical Literature 

ellipsis 147f. 

etymology 78; 81; 83f.; 109f. ; 122f.; 

Frauwallner, E . 136-138 
Gk)ldstucker, Th. 115; 142 
Greeks 120; 153 
Halil 79f. 

Hsuan-tsang 88 

internal reconstruction 109; 111 fn. 103 

Iranians 80 fn. 20; 89 

item and process 195 

I-tsing 154; 170 

Japanese 79; 167 fn. 22 

Jiun Sonja Onko 167 fn. 22 

Kielhom,E. 115; 135f. 

Korean 79 
macaronic style 184 
Middle Indo-Aryan 106; 116; 131f.; 
162; 169; 188 

Mimamsa 83 fn. 32; 110; 124; 159; 
175 fh. 30 

nouns 111 ; 128; 155 
Kyaya-Vaisesika 159; 175 fn. 30 
object 94f.; 97; 160; 173 
Pali 139; 194f. 

paradigms 163; 168; 174; 189; 194 
passive voice 97 f.; 155 
patronage 169; 187; 190; 196; 199 
periphrastic perfect 106; 144 

person 112; 199 

personal endings 96 f.; 155 f.; 163; 168 
fn. 4 

phonetics 78-80; 82; 112; 128; 130-132; 

176f.; 198f. 
ritual 80 

roots 84f.; 110; 196; 199 
Samkhya 159 

script 79; 113; 162; 171; 198f. 
substitution 96; 109; 116; 130; 140f.; 

suffixes 110 

symbols 109; 112; 131; 151; 168f.; 189 
syntax 94; 113; 181 
Tantrism 78; 92 fn. 24; 172; 175 
Thonmi Sambhota 167 fn. 22 
Tibetan 167 fn. 22 

topical arrangement 123; 129; 131; 158 
transfer grammar 159; 188; 192f.; 196f. 
Veda, Vedic 77f.; 88f.; 106-108; 117; 
122; 144; 164; 168; 174; 176; 187; 

verbal stems 85; 96; 107 fn. 89; 155f.; 

182; 196; 199 
verbs lllf.; 128; 155 
Whitney, W.D. 107; 166 
word and paradigm 195 
zero suffix 110